Good afternoon, everyone. We're about to get things started. The webcast is live. If you haven't already, you can download the presentations on our investor relations website. I think a few of you have been saying you've been able to do that. I was a little bit worried about the numbers doing it all at once, but I think we're okay. Again, welcome to The 2019 Trade Desk Investor Day. We're really pleased to have a super packed room here. The webcast, it looked like it was approaching over 1,000. There's a lot of people on today, and we're really thrilled to have that. Just a couple of points before we get rolling is I talked about where you can download the presentations. There's, of course, a live webcast that's going on through the Q&A.
In terms of Q&A, we ask that you refrain from any questions throughout the presentations. Then we'll have an open mic where myself and Jen Lee will be walking around the room, and you can address the executive team here at The Trade Desk. We also have two 10-minute breaks planned. There'll be time to roam around. We have food and drinks in the back. We have a full agenda where we have six speakers today. I guess before we start, I just have to give the forward-looking statements where please be aware that we will be making forward-looking statements in the presentations today. There's risk associated with that. You can refer to our SEC filings, which are on our website, in our press releases. That is all available at investors.thetradedesk.com. We've covered that.
With that, I will announce our Founder and CEO, Jeff Green. Jeff.
Thank you. Thanks everybody for being here. It's exciting to be back here. We're two and a half years in as a public company. It's been almost 18 months since we had our first Investor Day. So much has happened in the world and with our company and our stock since then. It's great. Yeah. It's been great to get to know so many of you and to have people who believed early in our company here and us all sort of benefiting and enjoying the ride that we've been on. I'm so excited to tell you about the future.
This is a really exciting time for me because I feel like in the last two and a half years, I've gotten to know so many of you, and this has become an extension of the presentation I gave a month ago to our own employees, where we're just talking about our future. Thanks for being a part of the team. Today I'm going to talk about our vision and strategy for growth, I want to start big. I believe that The Trade Desk fuels the world's growth. It's not just that we go where GDP growth is. We actually are also promoting GDP growth around the world. When we go to places like Asia, where there is so much opportunity in the world, it's experiencing the largest rise in middle class, arguably in the history of the world.
We are not just there to ride the wave, but to accelerate it, advertising does that. We're so excited to be a part of powering, for instance, nearly every major CPG company in the world, where they're helping people choose for the very first time what laundry detergent they'll use or what washing machine they'll put that in. It's just an exciting time to be a part of that. As I've said before, the difference between somebody like Dropbox being the winner of that cloud-based storage and the 100 other companies that were doing things like Dropbox 10 years ago, I believe is marketing, advertising, the ease of on-ramps. Without being good at advertising, it is more difficult than ever to be successful because of competition. Advertising is more of a difference maker than it has ever been.
We think that long term, in the same way that maybe Martin Sorrell was for the last 10 years, we can become a bellwether for the way the economy is going and who's doing well and who's not because we power so much of the world's economy. Advertising today is a $725 billion industry. When I first started telling these stories, in fact, I'll reference this, but when I first started talking about this company, my very first deck, I was pitching the size of the TAM, I was rounding up because that's what you do when you're in startup mode. The industry was at $580 billion, I was rounding up to $600 billion nine years ago. I just mentioned that to say we've gone from $580 billion to $725 billion, and of course, if you extrapolate that, we're a $1 trillion industry in eight years.
Many of you who were at our investor day last time. I'm sorry, I'm going to stand in the way. Many of you who were at our investor day 18 months ago, we said the exact same thing. In 10 years, we'll be at $1 trillion. That march continues, if you look at GDP growth, it's something like 150% in some cases, depending on where you're at in the world. Double GDP is where we see advertising investments being. Again, advertising is leading GDP growth. I just look at this as there's the $1 trillion TAM, there's the big circle, that's all of advertising. Now let's say what's in digital. This is where we're basically all looking at the secular tailwinds that obviously make up the investment thesis of so many people in the room.
digital just continues to get a bigger piece of the pie, and one thing I would encourage everybody to think about is that even in the cases where the ad itself is not digital, there is no reason why the transaction shouldn't be digital. Of course, you can put a print ad in the middle of Times Square, there's no reason why we have to sign that on a piece of paper just because the ad itself will be on paper. Same with magazine or whatever. Even though of course, that medium is shrinking in terms of total dollars because digital out of home will replace most of that.
Most people are reading journalism on a device and not in print, that doesn't change the fact that there will always be some amount of hard tangible advertising that isn't digital, but the transactions themselves will be digital. I actually think this understates how quickly we're going to go to everything being transacted digitally. You see that this secular tailwind is only getting stronger for us. Of course, programmatic is marching along with it. What tends to happen whenever the analysis is done to figure out what's going to happen in programmatic, the same thing with the erosion of traditional television, people tend to be very linear and they look at the trend lines, and it's very difficult to account for when things will hockey stick.
What are the major tipping points that make it so things accelerate or decelerate depending on the way that you're looking at it? I, again, think that this understates what's going to happen with programmatic because of the fact that programmatic is just better. It is better to transact with more data than it is to be guessing or doing things in the sort of prehistoric way. I think this understates what's happening with programmatic, but even if you just look at it from the linear perspective that I think most of these models suggest, again, a major secular tailwind and we continue to grow. I believe that eventually 100% of transactions or nearly 100% of transactions will be programmatic. You may say, "What won't be? Why nearly 100%?" Because whenever you're testing something new, you haven't necessarily standardized it yet so that it can be transacted programmatically.
As soon as you find something new that is good, you will standardize it so you can welcome as much supply and/or demand as possible, and that by its very nature, requires it to be where all the liquidity is, which is in programmatic pipes. Digital pipes too. Of course, there's by region. I just want to highlight, this is part of the reason why I spent most of last year living in Asia. It's the reason why I spend a lot of time talking about it, despite the fact that North America represents about 85% of our revenue today.
Of course, if you look at those secular tailwinds and you spend a lot of time like I do, thinking about where all the growth in the world is going to be, you cannot ignore the fact that the growth rates in the rest of the world are higher, as well as the fact that two-thirds of global advertising spend is outside of North America today, and that given that 85% of our revenue is here, we have some catching up to do in the rest of the world. This also just highlights where we think the opportunity is. We're going to continue to invest aggressively in the rest of the world, and it's one of the reasons why we think it's so exciting that outside the U.S. has grown at more than 150%, almost double what we've done in the United States since being public.
Let me talk about some of the macro things that have happened since our last investor day. A lot has happened. Let me highlight a couple of them and why they're significant for us. Sorry, I need water. This was a huge event. You may wonder why this is relevant to The Trade Desk. This changed everything in advertising and in our world. You may say, "Why?" I would just summarize it as, this was very uncomfortable to watch because it was a surprise at how little our legislators understood what Facebook is and what they do. It was hard for Mark to answer questions because he's obviously quite literally in the hot seat as they're asking really, in some cases, ridiculous questions.
The reason why this was a huge event for digital advertising is because everybody was watching, especially Google, said, "I never ever want to have to sit in that seat and answer those types of questions." It changed the strategies for Facebook and for Google. The fact that that was so uncomfortable made it so that they all became much more conservative in the way that they're handling data in particular, because I think both of those companies realized that they were taking unnecessary risks because of the personally identifiable information that they control, and especially the directly identifiable information that come to them on a daily basis that they had to be more careful with.
That meant that their lower priority, which is advertising to the rest of the world, meaning their number one priority at Google and Facebook is both to monetize Google and YouTube at Google. Their number one priority at Facebook is to monetize facebook.com and the Facebook app. It is not to monetize the rest of the internet. Because of this sensitivity, it made them both downgrade their desire to be on the rest of the internet, which opened an opportunity for us. Another macro event was that AppNexus was purchased by AT&T.
I spend a lot of time in earnings reports since our last investor day talking about how excited I am by AT&T's strategy to go all in on 5G, to go all in on on-demand content, to spend $100 billion effectively on content, and make it so that they can transform the cable experience to look more like television. Or, I'm sorry, to look more like Netflix, so that TV is a new experience. They publicly said that that needs to include programmatic advertising as a part of the future of TV. Of course, they bought AppNexus in one of the biggest deals that's been done in ad tech history, that repurposes AppNexus, in part, to focus on linear television, which is, I think, somewhat shortsighted, but I totally understand why AT&T would need to do that, given they paid $100 billion for all of that content.
What that does open up for us is one of our competitors in AppNexus, is now less of a competitor in international markets, as well as on the API front, which is something that we've gotten way more focused on. It also makes them more of a sell-side player, which they always were, sell-side and buy-side with a bias towards the sell-side. They become way more partner than they are competitor almost overnight, it becomes really exciting for what will happen to AT&T. This was a sort of win-win-win for us in the sense that we get opportunity in APIs, we get a great partner on the supply side, and we accelerate connected TV, which I think is the biggest opportunity we've ever seen and will ever see. OpenX is a company you probably don't know as well.
They're one of the three biggest SSPs. Incidentally, my notes are not on this screen, so whoever's controlling that, if one of the screens could transition to notes, that would be better. Anyway, I'll press on. OpenX. Very good. Thank you. The reason why this is so significant is because most SSPs were at a crossroads three or four years ago, where they had to choose, are we going to go compete with DFP or are we going to focus on being pipes? If you think about it the way I do, basically, there's buy-side and those are the brands and the agencies. Then there's the DSP, which is the buy-side tech. Then there's the exchange, which is not dissimilar from Nasdaq, where it's an unbiased pipes. Then there's the sell-side technology, where it's representing the publisher, then there's the publisher.
There's those five entities. The DSP is sort of like the prosecuting attorney, the SSP is like the defense attorney, then you've got the judge and the jury in the middle. Those pipes don't make, as a percentage, very much money. They make it up in volume. That is, of course, the way markets like this one, the Nasdaq, work. When you're at a crossroads, like OpenX was three or four years ago, deciding, do I go compete with DFP and try to make a greater percentage of every dollar? That's hard. DFP is humongous. DoubleClick for Publishers, which is owned by Google. Do I go compete with Google or do I just focus on being pipes? They decided to focus on being pipes the way that nearly every SSP did.
However, in the short term, they were charging still 20% or 30% for that function. OpenX doing layoffs is a great thing for our industry. It's of course not a great thing for the 100 people that were asked to go find a new job. The reason why this is so important is because I look at it as, you've got a supply chain with especially Google and Facebook, but from all walled gardens, where they're often operating at 50%-plus margins, but they control all of it themselves. The rest of us in the alternative to those, there's more variety, there's way more media, but there's also more moving parts. If somebody's taking more than their take or more than their share, more than they contribute, then the value proposition of the independent internet is lower than it would be even with really healthy margins here.
When competition works and when effective markets do their thing and that reduces, then they have to be very cost-conscious in order to stay competitive. This is the earmark of very good things happening in our space, which is it's putting pressure on companies to be more competitive, and that's exactly what OpenX is doing. I expect them to be around long term. They're a great company. They're well run. Their CEO is a fantastic person. They're going to do well. This represents a good thing for our industry. I think my favorite Christmas present of 2018, and it came in the summer, so it was early. My favorite gift of 2018 was Google removing their ID or announcing that they were removing their ID.
I've spent a little bit of time talking about this on earnings calls, but I just wanted to spend another minute talking about it today. Historically, Google would make their ID available to everybody, and they would give in their reports a user-level log. They would show, you showed this ad on this website, whether that was ours or on somebody else's, to this anonymized user. The thing that that enabled you to do is take their reports and give them to somebody independent, and you could compare apples to apples, meaning you could compare Google performance to others. Facebook didn't really do that. They, at times, had aspirations to do that when they bought aQuantive or I'm sorry, when they bought Atlas, not aQuantive. Microsoft had aQuantive. Atlas was purchased from Microsoft, so a piece of that.
When they bought Atlas, they were thinking of doing the same thing, they never did it. By Google removing that ID, which was all because of privacy concerns, I believe it was the result of that Zuckerberg hot seat that they made the decision to remove the ID, but it made it so nobody can really grade Google and compare them to the performance of the rest of the internet. It makes it so Google effectively is grading their own homework, necessarily. They have to. It makes it so that you don't get the same insight.
If you're a big brand, I've heard this from many big brands, they look at it as there is an asymmetry that has gotten worse, which is I give Google user-level data so that they can help me figure out where to target the users that I care most about, Google gives me back something less than that. They're also grading their own homework in terms of what I bought. Great, it's working, they're also then able to take credit for the entire sort of purchase funnel because all they see is Google, because Google is often the last touch, meaning, you advertise for 20 years about a Mercedes-Benz, you see commercials, then at the very end, you type in, "Go to Mercedes-Benz dealer," or "What's the closest Mercedes-Benz dealer?" You click on that ad.
It makes it so that in some cases, Google can take credit for everything that's happened above that. Because they're not giving you a mechanism to see what events were above that so you can do more complicated attribution, it leaves no conclusion other than Google touched it, I have to give them more credit for it. All of that is making it more difficult for a sophisticated brand to be holistic about their buying, it makes them look for alternatives. The reason why this is such a gift for us is because while they're moving to asymmetry and taking things away, because we don't transact in personally identifiable information or directly identifiable information for consumers, we're able to continue to give the ID in an anonymized way and continue to be symmetrical in the way that we transact with our clients.
As many of them have said, "I want an alternative to Google as a result of this," we're the first call that they make. I don't believe we've seen even the tip of the iceberg in terms of the benefit that we will see as a result of this policy change. Next thing is, of course, there's been a lot of media tug-of-wars that have happened, I don't know that there's any that was more significant than Disney buying Fox, which of course, makes it so that they have total control over Hulu. I basically, I'll talk more about this in a few minutes, I believe that most of television content has been at a crossroads over the last couple of years, where they're looking at Netflix with envy.
The very best thing that has happened since our last Investor Day in the television industry is actually the success of Amazon and Netflix because it's made all the traditional TV companies say, "We have to catch up. We have to make changes so that we're starting to look more and more like they are." That's why AT&T made a huge bet. It's why, of course, Time Warner and HBO are in that umbrella. I believe that's also why Hulu and Fox end up at Disney. The crossroads that they were at is they said, "Okay, we have to divide and conquer. We can't be all things to all people." All of them recognize that Netflix is ahead of them.
The crossroads that they have to choose between is, are we going to try to control distribution, or are we going to focus on the content and going direct to consumers? Are we going to source the demand, or are we going to focus on getting the content direct to consumers? What all of them have said, at least preliminarily, is they're going to focus on developing a relationship direct to consumers related to their content. That means that they're welcoming demand from us. We look at this as maybe the most exciting thing that we're seeing in the landscape today. There's no way that Hulu is going to be the same now that it's fully controlled by Disney, at least in my view.
I think that while it will have more Disney content than it's ever had and it will be a distribution channel for them, I fully expect us to continue to be one of their biggest programmatic partners, if not their biggest, for as far as we can see into the future. I mentioned I was going to share a number of things that happened since our last Investor Day on a macro scale. I'd like to talk about some of the things that have happened in our own four walls. The first is perhaps the most exciting, which is that in 2017, our revenue was 52%, and in 2018, our year-over-year revenue growth was at 55%.
I said 18 months ago at Investor Day that I believe that at some point in the next five years, revenue would accelerate for us and that our industry would accelerate. I did not expect when I said that it would happen so quickly. The single biggest explanation for that is actually connected TV, which I'll talk more about. We also, since our last Investor Day, did the biggest fraud prevention deal, I think, in advertising history. You probably didn't hear much about it. Part of that is a deliberate decision on our part to just not talk too much about fraud. It feels like there's more downside than upside when you throw a lot of word count to the word fraud. I wanted to take a minute to explain what we did there and why it's so significant.
There are many companies out there that what they do when they claim to be fraud prevention is they actually don't want fraud to end. They actually just want to tax it. They just go out to the world and just sell fear, but they don't ever want to totally eradicate fraud because that's how they exist. You see this with insurance companies or there are just so many places where there are business models that are dependent on not actually solving the problem, but sparking or fueling the fire around people's fear. Many companies in fraud prevention in our space did exactly that same thing. We were in this really unique position, which is because we're the buyer and we see the entire supply.
We see all 10 million ads per second, we also get to look at the fraud prevention techniques of all these various companies. We found a company that was performing meaningfully better than everyone else. We were afraid that what they would do is what everybody else had done, which is raise too much money and then have to have a sort of go-to-market strategy. Which is to sell fear to everybody and then sell your product to the publisher, to the SSP, to the exchange, to the ad server, and sell it all along the way. You basically sell it to everyone so that you can tax it as much as possible. Then again, the independent internet on this side takes a big hit versus the walled gardens when it comes to the efficiency of things like fraud prevention.
We took a very proactive role here, and we went to the company and said, "Hey, we think you're doing this as well as anybody. Don't raise any more money. Let us be your distribution. Let's implement this right in the industry so that we can actually stop fraud." The type of fraud that we're talking about is the one that is, I think, the most pervasive and the biggest problem, or was the most pervasive and is the biggest problem, which is distributed non-human, where something, they put a virus on computers, and then it's non-humans basically simulating the actions of human beings so that you can get false CPMs and get money from advertisers.
One of the problems with many of the companies, if you don't get implemented everywhere, it's sort of like fumigating two-thirds of your house, then the bugs will all go to the other third. It's a ridiculous way to operate, but that is mostly what happens in our industry, is that if you clean it up in one place, it goes everywhere. It became a place where we said, I believe Google has more at stake than anybody on this. We should work closely with them because the two of us want to eradicate fraud, and we're fine to use an independent company. We worked very closely together, and we said, "Okay, here's the supply chain where they typically tax everywhere in the supply chain. Let's go to the very top. We'll implement this directly at the SSP level.
We'll require them to implement it, and we'll require them to pass the cost on." We actually called a meeting with 20 people in our industry who are the CEOs of all the biggest SSPs, including with the fraud prevention company, which is called White Ops. We, in advance of that meeting, asked White Ops to cut their costs, what they charge their customers by more than half. We worked with them to develop a new revenue model. Essentially, what we were saying to the SSPs is that only if we cut the cost, can we ask you to implement this across the board. We basically told them that we would stop buying at the scale we were if they didn't implement it everywhere.
Google did the same thing, as a result of both of us sort of pushing this on every major supply source on the internet, we made it so that we fumigated the entire house, and you spend way less time talking about fraud than we ever have. Something that all happened behind the scenes, but it's something that we've done to make it so that the supply chain is more effective, and it's why I believe all of us can have more confidence in this independent internet than we've ever had before because of things like this. We also acquired Adbrain, and we estimate that we've made at least a 20x return so far. If I were just using direct ROI, it would be more like 8x. But when I say direct ROI, I mean revenue that we've made for charging for that specific feature set.
What Adbrain does, if any of you are unaware, is one of the biggest problems that happens in advertising today, and this is especially true in the walled gardens, is that you'll have one experience in the walled garden or on one site, and then you'll go see the same ad on a different site or on a different device. You see five ads on your desktop computer. You still meet their targeting criteria when you go to your mobile phone, they're going to hit you again there. Unless somebody does the work, the mathematical and statistical work to weave together that this computer is operated by the same person as this mobile phone or this connected TV or this laptop, then you will never be coordinated in the way that you advertise across the entire ecosystem.
Adbrain does that mathematical and statistical work to weave those together. We believe they were one of the best, if not the best, in market at doing that very focused thing, but they had not reached scale, and we were their biggest customer, so we were able to get a phenomenal deal on that company. Because of not just the 8x return that we've made on what we charge for their service, we've also been able to get more revenue as a result of having them on board. We've done one acquisition in our company history, and we knocked it out of the park in terms of ROI. You'll remember that we introduced a new product set called Next Wave in the middle of next year.
One piece of that was a new user experience, where we totally transformed the way people engage with our UI. Maybe many of you, whether it's as a consumer or as an investor, lived through Microsoft doing that well or not well at times. There are sometimes you got a new version of Windows you're excited about, sometimes you're frustrated by it. Giving a new user experience, even though we were convinced that it was a massive upgrade, you always worry about what the adoption of your users is going to look like, especially because there are just so many use cases. There's so many different reasons that people use it. Did we think about everything? Are we going to alienate any of our user base? We had all those anxieties for, I think, all the right reasons.
It is, without a doubt, it was the most expensive product release we've ever done, it is also the most successful product release we've ever had. The fact that 75% of our clients are using that new user experience, even without it having 100% of the feature set, because today we still support both versions, so you can have the old version or the new version. The fact that it's only been since June and we have already 75% adoption, especially when you consider that there's an annual planning cycle and many clients are using both today, but they've actually done the extra work to move people over faster. It's just an unbelievable or incredible success. I'm so pleased with our product team, our product marketing team, our marketing team, our sales team, our client-facing teams for getting that sort of adoption.
I've used statements before like, "The most bullish number I think I've ever shared." I think I've said that phrase three or four times on earnings reports because we just keep replacing the previous most bullish statement that we've ever made with updated most bullish statements we've ever made. Maybe those are all incorporated in this CAGR of 600% spend since 2016. Of course, the first year at 1,000% is impressive, but the numbers are small. I actually think the numbers that we put up last year are the most impressive, just because a third year in a row of over 1,000% on specific types of spend is especially impressive. Also on the micro, I just want to talk about a case study with TVB. I mentioned that I lived in Hong Kong for most of last year.
It was partly to just learn the landscape and understand the culture. It was really for me to learn a lot and then hire people to make it so that we could scale more aggressively. It was amazingly successful, but one bonus that came from that was a relationship that I formed with TVB, which I think we created one of the most impressive case studies for connected TV anywhere in the world. The reason why it's so impressive is because TVB in a city not too much smaller than New York, Hong Kong is unique in that it's really the only major city in the world where everybody speaks Cantonese. It's not unexpected, or it's not surprising that TVB is kind of like NBC, Fox, and ABC rolled into one, where you've got the major network. They're very data-driven.
TVB said, "Well, if we move to everything being OTT, would that actually save us money?" Because of the fact that all the content is in Cantonese, they felt like they needed to move faster in order to compete with Netflix, because they're much smaller than the footprint of the major media companies in the U.S. By running the math, as well as making some strategic decisions about the way trends looked from international expansion plans from Netflix, they said, "Let's launch our own device that is sort of like Roku. It'll be built on top of Android.
We'll ship that to as many households as we possibly can in Hong Kong and see where that takes us." I was there for their two-year anniversary, where they opened the celebration by saying, "We're delighted to have our programmatic partner in The Trade Desk here." They basically outlined how they went all in on both programmatic, digital, and OTT. Using that device that is like a Roku, they got 60% of Hong Kong citizens in less than two years to insert the box, pay a $100 fee, and basically get all their content on demand going forward. They got higher CPMs. It was a massive success by going all in.
The reason why I think that this is one of the best case studies, maybe the best case study in the world on the adoption of OTT, is that many of the bigger media companies in the U.S. and also around the rest of the world have said, "Well, our cash cow is still linear TV. Do we want to go all in on this? Should we be afraid of going all in?" Being very analytical about it and also strategic, they went all in and proved to be massively successful by doing that, and are also ahead of the curve of what's inevitable, which is consumers demanding on-demand content. TVB is a massive success for us, and we're super excited to be their primary, or at least one of the leading programmatic demand sources. I think we're the primary one.
Since 2015, our non-U.S. revenue has grown from 6.6% to 14% in 2018, which basically underscores what I've said, which is that international spend outside the United States, is expected to grow somewhere in the ballpark of double what we expect the U.S. to grow. We are over-investing around the world everywhere, Europe and Asia in particular. I'm super excited by that number and that we continue to grow internationally at a faster pace. I actually hate using the phrase international, by the way. I just haven't figured out a better way to refer to it, because it's not like we're a U.S. company. We are a global company that happens to be based in the United States. Because we think at end state, two-thirds of our revenue or more will come from outside the United States.
I say two-thirds because right now, two-thirds of the advertising pie is outside the U.S. Because it follows GDP, that means that because there's more GDP growth in Asia, more than 67% will be outside of North America. I just want to underscore that we're a global company that happens to be based in the United States. We're not a U.S. company. I want to share with you just a little bit of the success that's come from Next Wave, and rather than me talking about it, I just wanted to give you a minute or two video of exactly what it is and what we've done. We are on a mission to change advertising, and today we are reinventing media buying.
The Next Wave is really about enhancing your data-driven decisioning, helping you to make smarter, faster, easier data-informed decisions while still losing none of the control or transparency that you're used to.
The new launch means streamlining the programmatic process from start to finish. We're now going to have strategy intertwined with traders, making everything more seamless and improving campaign performance.
I think advertisers can now make smart and insightful decisions. This means greater efficiency and lesser time and money spent on testing and learning.
The Next Wave is three main products. Planner, which is a new data-driven planning tool. Megagon, which is a new and intuitive trading interface, and Koa, The Trade Desk's artificial intelligence that powers all of this. It's hard to give enough words to what we put into that. You may recall that we put 40% of our engineering resources for almost two years into building those three products, the fact that the adoption has been so high as some of those stats at the end highlighted, has been a massive success for us. We'll use that as a template for big releases going forward, which is something that I'm especially excited about. I want to spend a minute or at least a few minutes actually talking about our culture. I'll talk mostly about this near the end of the presentation.
I just want all of you to know that we've done something different than what other so-called unicorns have done which is we've optimized to our culture. We've let that be the bottleneck. For many of the companies in Silicon Valley, they've just said, "Oh, we just hire. You look at the financial statements. You can reinvest. Let's go as aggressively as we possibly can. That's what we need to do in order to sustain growth." You look at revenue per employee, then you say, "Well, then we just need more employees, then they can go do all these things." What happens is you don't really realize the effect that you're having on culture. We've been okay letting our culture be the thing that slows us down. I'm not willing to mess that up because you can never really fix it.
I think you can improve it, but it will never be what it once was. There's something about protecting that we're just more committed to than we've ever been. This is a picture taken. We have our annual get-together, our family reunion, if you will. We call it Palooza, we get all 25 offices. We fly them all to be together for a week. We are very good at having a common vision, we're good at making different cultures from around the world work together. We collaborate really well. We love being together. We truly celebrate our diversity instead of being afraid of it. You'll hear from Brian, who's on guitar right here. You'll hear from Dave, our CTO, who's on bass right here. This is one of our office managers in Ventura.
It's all of us getting together and just spending some quality time together, mostly focused on vision and understanding exactly what we're after and what we need to do to be laser-focused, but it's also having some fun. I mentioned that we have offices all over the world. We definitely have done a great job of representing California in our office distribution. That's a commentary on traffic. Same with in Washington. Those bridges are brutal. We continue to expand around the world and you'll continue to see us add to those dots. If you look at the biggest media companies in the world, you see that we've represented nearly all of the top 15 major markets. Of course, we've not forgotten that we need to get into India and one or two of those that we've already passed over.
I'll also mention that Singapore ends up supporting a lot of those fast-growing ones as well. That's initially how we got into Indonesia, for instance, but we're already supporting places like Vietnam, which are super fast-growing, from Singapore. I mentioned our culture. We were really honored to be recognized as Fortune Magazine's number 2 best places to work in the U.S. for technology companies. We work really hard to protect our culture. These are three of the, I don't know, 25 or 30 awards that we've won since we were last here. I don't put too much stock into them. It's nice to be recognized, but I do think that the culture at The Trade Desk is one of the things that is most exciting about our future, and we just don't get enough opportunity to articulate it.
I still will forever feel like I'm falling short of describing what a unique and special place this is. At some point, I should probably share employee retention numbers and some of those things just because we've done such an amazing job of that. It is part of the reason for our success, that we focus on people in a way that I don't think most companies do. All right. Let's talk about the future. Before I do that, I need another drink of water. As I've said before, the Mad Men era is over, so water. Many of you may be asking, why are we leading? Why is The Trade Desk doing so well? I have said this before publicly. I loved this interview that was done with Mark Zuckerberg five or six years ago, where somebody asked him why Facebook won.
Why didn't Microsoft or somebody else develop the world's largest social network? His response was, "They should've. We won because we cared more than they did. We were just laser-focused on this one thing. We cared more." I believe that's exactly what's happening with objective media buying. The reason why we're doing really well is because we are not trying to do a whole bunch of things. We're just trying to do one thing really well, which is help people objectively buy media. We care about our employees, and that's why-- Initially, on this slide, I wrote, instead of our employees matter, it's our employees believe they matter. Of course, they matter. Oftentimes, employees don't believe they matter.
If you work in big companies, it's one of the hardest things to do, is to convince all 90,000 or whatever number of employees that you have, that they all matter. The toughest part at working at a big company in Silicon Valley as an engineer is that as you're writing code, in the back of your mind, you're thinking, this will probably never see the light of day. When an engineer at our company is writing code, they're saying, "Oh my God, I've got to get this out by Monday because this can make a huge impact." I think one of the things we've done really well is we've developed systems and process to make it so that even at scale, especially on the engineering side, but across our entire company, everybody matters. Everybody's expected to contribute.
It's actually the hardest part of hiring because there was once a time that we were this unknown company where you had to have vision to join us. Now we're a place where people view us as safe. We have to work really hard to make certain that we don't get the people who are freeloading at those big companies to come over and then not contribute here. We have to work really hard to make certain that not only do you matter, but you matter. We need you to do this. The double-edged sword of mattering is something we spend a lot of time talking about. In order to do that well, you have to, of course, convince them that you care. We care not just about our employees, of course, but we also care about our clients.
I think that's one of the big reasons why we've been successful is, and I think this in general, which is the most successful companies of the new era are those that have done a better job at aligning their interests with their clients than the sort of incumbent. That's exactly what we've done with the ad agencies. If you go to a tech conference in Silicon Valley, it's called Disrupt, right? The average company in Silicon Valley is trying to disrupt the status quo. To use an example, Uber, to some extent, is trying to make the medallion of a taxi driver worth nothing. They're doing well. They're disrupting the status quo, and that means that somebody who has historically done well is going to lose. We're not.
We're an enabler, if you will, and we're going to ad agencies, and we're saying, "You matter." Because you have developed operating leverage, operating leverage for you and for us. WPP has over 100,000 employees. The only way to get that big in advertising is to do what they've done, which is acquire companies for three decades. Instead of us going and saying, "We're going to disrupt that," we instead say, "Let us help you transform your business, and you have to, or you're going to lose. You have to transition." Unlike most tech companies who are trying to disrupt all the people that have historically been the winners in the space, we're going in saying, "Let us help you.
Let us enable you transition into the digital era and be better. That makes it so instead of a huge group of people rooting against you have almost everybody rooting for you in being successful. As long as we can continue to maintain our humility, which we work really hard to do, I think we can continue to get them to support us. By them, I mean our clients, which none of this is possible without them, and they deserve as much credit as anybody for our success. We're a SaaS-like company that's pointed at data and advertising. We're not a media company. I just want to remind everybody, we have 95%-plus client retention, and we have a more profitable business model than traditional SaaS.
I think it's really important to think about what is most important about SaaS, because that's where we've spent a lot of time obsessing. For us, it's about client retention, profitability, and predictability. It's not about subscription. A lot of times people are hung up by the fact that we are not a subscription service. When you have those other three things, and especially if you make more money than you would if you were on a subscription, I believe more SaaS companies should be using revenue models like ours. Instead of going to your clients and saying, "Hey, I'd like to get married on a first date and I need a four-year commitment," which is effectively the way that they sell, you can go in and say, "I'm confident our product is good. You can cancel anytime you want.
I am so committed to consumer surplus, to making our product good, that we're confident that you won't leave. I'm willing to put all of my effort and money where my mouth is." That's what we've done time after time. Despite all that, despite the easy out, we still have 95%-plus client retention. I think we've managed to get the best of both worlds. Every one of these, incidentally, at our last Investor Day, it was a different picture, but the bullets were exactly the same. I wrote these bullets 8 years ago. I just want to remind you of the bullets because nothing in our core beliefs, nothing in our value proposition has changed substantially since then. The biggest problem with the soon-to-be $1 trillion industry in advertising is that there's no price discovery of any kind outside of programmatic.
Price discovery is weak. People don't know what they're buying and selling. They know that it works. They just don't know which part works. There is no way to solve that without programmatic, and I believe that it's that fact that makes it so it is unstoppable. There is no single company in the world that can stop that from happening. There's no government that can stop that from happening because $1 trillion is too big to stop. Data-driven is just better than guessing. That's just a fact. Programmatic will eventually power all, and again, I wrote these 8 years ago, all 600. Like I said, I rounded up. It was five and change. The buy side is forever in the power position.
This comes from the fact that because the sell side is fragmenting, the bullet at the bottom on the left there, what that means is that as it fragments, it means you have to plug into more sources. It means that I look at what today is 10 million QPS. When I wrote this, it was measured in the hundreds of thousands. I get to look at all 10 million QPS, but on the sell side, they look at a fragment of it. The data asymmetry between buy side and sell side is humongous, and that creates a power position as it relates to price discovery for the buy side. Especially as TV comes online because people are protecting their content, there will be more fragmentation. We will have to do more integrations.
That creates more barrier to entry, that creates more asymmetry between the buy side and sell side. This position that we've maintained for the last decade, that the buy side is in the power position, will continue to happen. We took a guess at saying we think the sum of the tech take rate is between 25% and 30% and that the buy side is going to get roughly two-thirds of that.
We think that continues to be true, as I know many of you have watched carefully what's happening on that side, and that as it stays in that range, it will stay ahead of the walled gardens in terms of what it takes out of the middle and the value proposition of the independent internet will out-offer or out-compete with any walled garden, which is why I believe at Endstate there are no walled gardens. On the buy side because of that asymmetry and because when you buy something, you learn more about it, you get more data, you get all the feedback, you get to find out what the performance actually was. When you actually buy it, you get more insight. That means that there's a natural tendency to look for economies of scale on the buy side that do not exist on the sell side.
As a result, we think that there's going to be three to 10 players on the buy side. At least some of them, at least one of them has to be an independent player. I think right now, Google's doing the best as the non-objective or non-independent player, and we're doing the best as the independent player. The rest of the seats right now are empty. To that end, I just want to be super clear because sometimes people assume that we're competing with Google and Facebook, and then they say, "Whoa, that's big. That seems scary." I don't think it is. Let me explain why. Dramatic pause. I just need water. The reason why is we're not competing with Google's core business. We have not built a UGC video platform. We have not built a search engine. We've not built a space program.
You're welcome. We're not building self-driving cars. We haven't bought a VR company. We are just trying to be really good at buying media objectively on the rest of the internet. I have sometimes sort of facetiously said we are competing with the 47th highest priority at Google. Now I jokingly say, now we're competing with the 58th highest priority at Google because when they removed that ID, they made a conscious decision to move down the priority of monetizing the rest of the internet. I think the same thing is true at Facebook. I can't remember where we were at the Atlas strategy at last Investor Day, but three or four years ago, they were focused on trying to monetize the rest of the internet, and they basically abandoned that when they gave up their efforts on Atlas.
Monetizing non-owned and operated sites or destinations is not the priority of these companies, and that's what we're focused on. We believe that we're a very critical part of the future of connected TV, as especially the ad-funded options are trying to figure out how to get higher CPMs to pay for all their content. We're really excited about all the partners on here and thousands of others. I've said this basically since we've been public. I've said it years before then as well, but it continues to be more and more true. Objectivity is mattering more and more as time goes on. I don't have a dog in the hunt in the sense that I'm not just trying to sell our owned and operated inventory because I don't have any. It doesn't matter to us whether we choose Spotify or The New York Times.
It was really exciting for me a few months ago, to sit with one of the Fortune 100 CMOs who is, to me, one of the most respectable CMOs in the world and saying, "We built our entire business model so that we could sit in front of you today and tell you that we're objective. You can trust us. We have built a business model that is worthy of your trust so that we can be trusted with what we think is your most prized asset, which is your spend, and your second most prized asset, which is your data.
With both of those things, we can do amazing things together, and we would never be able to do that if we hadn't built a business model that was objective." I don't think we could have ever been even in that room if we hadn't built a business model that didn't have media. Maybe this is the biggest question that we've been asked by investors recently. Chris told me in advance, this is the one that you can't fuck up. Anyway, with that as preface. Amazon's built an amazing business in advertising recently, so I get asked about it all the time. And I want to explain what I think they've done really well first, which is they've done a great job of taking their site and figuring out opportunities to monetize. And I think they've learned from what Facebook did.
You'll remember before Facebook went public, they didn't care much about ads, and then they figured out a way to not be too disruptive to the experience. All of us as consumers can debate whether they did that well or not, but as investors, we can all say they did that really well. Amazon, I think, is doing something very similar, which is they're monetizing their site. Nearly all of the money they've made in advertising is from them monetizing their own site.
I do think they have aspirations to monetize the rest of the Internet, but it's a very small initiative and a very small contribution to the huge amount of success that they've had in a short amount of time for monetizing their own site. Here's what I think is the most important thing to think about when you're thinking about Amazon, and I've said this to Amazon, as I expect them to be a close partner of ours, the same way that Google is, the same way that Baidu, Alibaba, or Tencent are. We have complicated relationships with all of those companies, but we become more and more trusted partners as time goes on. I expect that to happen with Amazon as well because of the challenge that they have in monetizing the rest of the Internet.
If you look at this from a CPG's perspective, where they're selling more of their product on Amazon than they ever did at Walmart, sort of pre-internet, pre-e-commerce. To them, Amazon is scarier than Walmart ever was. Walmart had a huge amount of negotiating leverage pre-internet. They still do, of course. Also, Amazon is asking for more of their advertising budget now as they're sort of the new end cap, and I would summarize all of Amazon's success so far as being the new end cap. They've done a phenomenal job of that. If you then say, "Now, can you give me all of your media budget?" Again, if you're looking at this through the lens of a CPG, they also sell a third of all baby wipes in America. They use white label products, right?
They also sell their own products alongside those. It seems like the new end cap is a lot easier to move around than the old one. With the power that Amazon has created, I think they have a bigger objectivity problem than any company in the world as it relates to going to advertisers and saying, "Trust me with your money, and trust me with your data." Because of that, I think strategically long term, it will be obvious that they should continue to do what they're doing, which is monetizing their site, and they'll try to do bigger things like what I think Google and Facebook have done well, which is creating marketplaces. When they do that, I think that we'll always be a part of the demand or providing demand for them.
That's where I expect our strong partnership to be in the future. Chris looks relieved, by the way. That seems good. Last year or at last Investor Day, that was a year and a half ago, we talked about our five goals over the next five-ish years, there's no hard time limit on any of these. I wanted to just give a little bit of color. I wanted to review them. They haven't changed. I wanted to just give a little bit of commentary on how we've made progress on them and how we're pacing. CTV and video will become our largest channel. I believe when you add up all of video, when we're at that $1 trillion mark, it will be ballpark half of the pie.
When you add up mobile video, when you add up short-form UGC, when you add up the premium video, when you add up OTT, when you put it all together, it's going to roughly be half of that $1 trillion pie. That represents the biggest opportunity that we'll ever see, and it will become our largest channel. As I've said before, I think the most bullish things that we've ever shared are numbers that show that we're getting way more than our share in this fast-growing space. China will become a top-three market for us. I continue to believe that will be true. I think everything is going in our favor, especially Baidu, Alibaba, and Tencent, who own more of the China market than Google, Facebook, and Amazon do in this market, as they have aspirations for growth.
It becomes increasingly important that they get demand from outside of China. That's where I think we will be a way more important partner to them. I think The Trade Desk is more important to Tencent's success than The Trade Desk is to Google's success or to Facebook's success, which is why I'm super optimistic about our partnerships there. I think there's a lot of things going for us in China. Incidentally, on March 26th, our product will become generally available for our outside-in strategy, which is for advertisers around the world to take incremental dollars into China. We're only weeks away from launching our first generally available product in China, where we'll start to get, as I said in earnings and then had to clarify, real spend. By that, I don't mean material and numbers.
I mean actual campaigns running in China after multiple years of building infrastructure on the other side of the wall, creating entities, hiring a double-digit number of people in China, me spending time in Hong Kong to hire regional leadership. Now, with all of that in place, we're ready to start adding incremental dollars into China. Super excited. This is probably the most complicated thing I can talk about today. Maybe the fraud prevention deal is its rival, but we said that we would build an ID footprint that would be larger than any one company. Our Adbrain acquisition was a huge part of that. They're basically bringing the brain or the math to the equation, but the data inputs come as just regular course of business.
One thing that I think most people don't fully appreciate, including many investors, is the fact that the rest of the Internet is really, really big. When you put together all the time that gets spent outside of the Internet, including when you start a portal, whether that is Facebook or somebody else, then go spend time in the rest of the Internet, the rest of media is really, really big. Facebook reaches 2.2 billion people. I think these are on an annual basis. YouTube, 1.8 billion. Again, dramatic pause. We are reaching 3.5 billion people by looking at those 10 million ads every single second and having an opportunity to buy media from literally millions of places and not just one destination. This is absolutely apples to oranges. I want to be super clear.
It is an amazing achievement to go get 2.2 billion consumers to sign up for your platform and then use that on a regular basis. We don't have as high a bar. We don't have to do that. We don't have to get them to sign up for anything. Instead, what we do is we'll partner with whichever publishers and media companies are getting attention from consumers. By doing that, we're everywhere without having to go win them. That's a feature, not a bug. The apples to oranges doesn't take anything away from what they've accomplished. By having access to the rest of the Internet, we make it possible to not only monetize the rest of it, but also to get the reach for advertisers that they really want and need. Data will grow at double the pace of media.
In 2018, data grew at 1.6 times. In the second half, data grew at two times that. In other words, in the second half of 2018, we kept our promise. It is 100% a byproduct of building a more data-driven UI so that we make it obvious for people to use data. It is better to use data than not to. I still believe we've merely scratched the surface. I think we said something like two point something data elements are used on every impression. I believe that should be over 100 on every impression as we get towards end state. We've not come even close to what data will be. If you think about it from a consumer standpoint, the amount of data exhaust that you're throwing off as a consumer has gone up exponentially over the last decade. The use of that hasn't.
Even though companies are getting better at it, they're just not using it yet because it's hard. You've got to do a lot of work, and especially great companies are trying to figure out how to be privacy-centric, how to do the right thing, and also show more relevant ads and content. I also said that we'll change media planning. We shipped our media Planner in the middle of last year. We still have a lot of work to do to change planning. Media planning is sort of the upstream problem with advertising. What I mean by that is, the way that a media plan is put together today is using mostly bad data, which is all guessing, and it's all because the workflows are so hardened that people use bad data to make wild guesses.
Then digital is assigned a budget at the end of that process. Majority of the time, that is still the way that it works. That at the end of the process, we get assigned a budget. That makes no sense because digital is where all the data exists. You should actually start with digital and then inform the rest of it. Honestly, I believe The Trade Desk should get the first dollar because of our objectivity, because we have so much reach, and because of what we're doing in connected TV. The Trade Desk should get the first dollar on every media plan. Then the walled garden should get the leftovers. Then after that, the rest of media should get the leftovers. I mean that because of data-driven and objectivity.
I do think over time, that is exactly what will happen if we continue to do our job well and we continue to enhance our product. We've got a lot of work to do on that, but I believe it is the right thing for our customers to do. This is the most exciting number that Dave shared at our Palooza. Dave was the one who ran the math to figure this out, which is that because of the efficiencies that we produced as well as we've 4x'd our engineering team since the bulk of the work was done on Next Wave. Because of 4x'ing that and then because of operational efficiencies, we have this year the bandwidth to do eight Next Waves. If you're thinking like I am, how do I distance ourselves from the closest competition in the rearview mirror?
Innovating is the only way to do it. We have developed the resources and the focus to do that. This is one of the most exciting things that I can share with you, that we are expanding the moat. I've said this a few times already in this presentation, at the end of the day, this is a tug-of-war between data-driven and guessing. It's not old school versus new school. It's not old media versus new media. It's not digital versus traditional. It is data-driven versus guessing. We're fine to show ads in print. They should be transacted in a data-driven way, not over martini lunches and uninformed bets, because that's where so much of advertising is still done. The majority of advertising is still bought in an uninformed way, and that represents a huge opportunity for us.
I've mentioned before that we're protecting and scaling our culture. I want to talk about something that we did at the end of last year to just give a little bit of visibility in how we've changed. This was our picture. We take a family photo every year as we all get together, and we're, I think, right at 1,000 as we took this picture just a couple weeks ago. In October, I walked our company through a substantial change to our org structure that we called Alinea. Actually, to go back for a second, we did a bunch of surveys, both formal and informal, of people about Palooza. The presentation that I give at Palooza is the one I get the most nervous for every year because I care so much about what they think.
I also am paying millions of dollars to fly all these people out to rally around a vision. If I screw that up, not only did I weaken our chances of hitting our goals for the year, I also wasted millions of dollars to do it. I get nervous before every presentation at Palooza. This year, the feedback was that it was our best Palooza yet, which incidentally, immediately frightens me, because now the bar is so high, we have a tough comp for next year. We were trying to dissect why this year's was the best that we'd ever had. I think it was because it was the first year, especially since being public, where our tone to the team wasn't, "You have no idea how lucky you are. You have no idea the legacy you've inherited.
We've worked so hard to get to this point. Please don't screw it up. All of you that are new, most of you are new, most of you haven't been here that long, don't screw it up. This is an amazing place. Keep it that way." That was our tone in years past. This year, our tone was, "It's time to get big." That's what we talked about with this. This is Alinea. That's the symbol for Alinea. You use it to start a new paragraph. It literally means a new line of thought, a new way of thinking. What we did, the changes aren't really drastic, they are very important course corrections. We implemented what I think is the most innovative career path that I've ever seen a company implement.
I wish I'd had it in the companies that I worked for before The Trade Desk, which was, we helped every employee identify whether they wanted to be what we call an agile manager, where you're only going to manage a few people, but you have the ability to sort of be a SWAT team and go accomplish amazing things with a small group of people. If they wanted to long-term be an individual contributor, and one of the big mistakes that I think big companies make, especially with really talented and smart people, is they take talented individual contributors and force them to become managers, which they don't want to do, and they're not good at. You can be senior at our company and not manage people.
We're totally fine with that, and we have to develop a path for those people so that we don't lose them. They have too much esoteric knowledge in their brains that if we lose them simply because we did a shitty job of career pathing, then we've made a big mistake. The same thing with managers. We want to make certain that if you're on the management track, that we make it so you're going to be successful. We excitingly said we're going to send every employee to Harvard. We invited them all to take a management training course. It's a rigorous course that Harvard offers in management training. We have too many managers that are first-time managers that I need them to develop wisdom at twice the pace that they would have if we didn't.
How can I just make them learn as fast as possible and make them great managers? Because as you all know, and many of you, if not all of you, have lived, you join a company and you leave a manager. If we can make our managers great, we can minimize attrition and take all that esoteric knowledge that they would be walking out the door with and keep it in the company so that it's super productive. We did a reorg to make it so that it's more regional. One of the things that you have a tendency to do when you first get started is centralize everything so that you can have control.
Over time, you quickly make it so, especially if you have 25 offices around the world, you quickly make it so that you have lots of employees who have to support every time zone in the world. You're going to make it so that they're up all the time, or what's even more likely, is they neglect regions of the world. It's super easy for our APAC team to say, "I have a question." "We'll get back to you on Thursday." You cannot grow like we need to in Asia if that's the model that you have. I'm offering visibility into all this so that you know that we're committed to continue to grow, and we're making all the changes that you need to in order to be a big, grown-up company, while at the same time protecting our culture.
If you hear me getting passionate about that's because this is where I've spent most of my time. Three years ago, I was spending lots of time with clients, spending a lot of time helping them evolve. Now, I'm spending a lot of time on this because this is what's going to make us grow to a $100 billion company. The way that I open the conversation with our employees is it took us nine years to go from zero to 1,000 employees. We need to go from 1,000 to 2,000 in less than three. The only way that we do that is if we change the way we think, if we change the way that we think about scale and growth. That's what we're doing. Let me talk about our investment strategy. Good segue to that.
There's no way to talk about our investments if you don't talk about CTV. I use this slide. Okay, I've got seven minutes. I use this slide to talk to the team at Palooza. When I said we just talked about a whole bunch of goals. We got to go into China. We got to get more Megagon adoption. We've got to introduce new UIs. We're going to improve measurement. We're going to improve our data products. It can be really easy to say, "Oh, my God, I've got to knock down all 10 pins. How do I do that?" You focus on the kingpin. In our case, the two things that we need to do really well this year to be successful are hiring and most importantly, get connected TV right.
The great parts of traditional TV have been that traditional TV has reach, and it has a common experience. We all watch the same Super Bowl ad at the same time. It has amazing scale. It also has amazing efficacy. The new reality TV, and by reality TV, I mean OTT, what is data-driven TV, the new reality in television is that there is lots of reach, and there's lots of advertisers, lots of investors, lots of agencies that don't understand how in the last two years, reach has changed dramatically in OTT. The reason why that may be news to those groups of people, both investors and advertisers, is that most of you are early adopters. You've been using OTT for a long time, and you also knew from looking at the numbers that it's been really small.
You early adopted, you talked about it at parties, you told your friends at the holidays, and they all have adopted it. It's going much faster than anybody predicted, including us. Instead of the common experience, it's a customized experience or on-demand experience, and that is critical to this all working. Netflix has changed the game by investing ahead, but they've obviously used a lot of cash to do that, and they've made it so that traditional companies need to figure out a way to make more money so that they can afford to compete on the quality of content for premium content. That means they have to get higher CPMs. I believe, and most of them believe this now, too, the only way they can do that is with higher CPMs.
I believe Randall Stephenson made more than $100 billion bet on that, I think actually you can include the investment he's making in 5G on that bet, that without fewer ads, they're not going to be able to recoup the investments that they made, especially in the media investments. I believe they will be successful at this. Whether or not they'll execute on every piece of it doesn't matter to us or to the ecosystem. The fact that everybody is going to go all in on the customized experience in search of higher CPMs so that you can fund premium content is the only way for the new reality in TV to be sustainable. Just a reminder of what's happened. Our 2017 inventory was up 1,000% December over December. I said at that point that was the most bullish number that we've ever shared.
I replaced that when in Q2, we said Q1 2018 over Q1 2017, our spend, not our inventory, but our spend was up 21x. That's the most exciting number I think I've ever shared. CTV availability 2018 over 2017 was up 6x, while our spend was up 9x over the same period. In other words, we're getting way more than our share, and when you do that multiple years in a row, obviously the numbers are getting material. CTV spend is material. China spend is not material yet, even though we're getting real campaigns running through it, but this is real campaigns and material spend. Data-driven OTT may be the best scaled advertising that I think we've ever seen.
The thing I want to just highlight is that you can make the argument that the bottom of funnel activities that are credited for being advertising are not really advertising. When you type in, go to Mercedes-Benz dealership and you see a text link, is that advertising? I think it's a philosophical question that is worth spending time on. It doesn't mean it's not valuable. It's super valuable, but maybe it's better categorized as navigation than it is advertising. I think advertising is about winning hearts and minds. That is done way before somebody types in buy the thing. When you're winning hearts and minds, I don't think there's ever been anything that's more effective at scale than television, where you have moving pictures and music. It's been very effective, and it's been the lion's share of advertising for a long time.
Now we're actually turbocharging that by adding data to not just who sees it, but also long-term for the way the content itself is created and customized. It will be way more effective than television's ever been, and I would argue that at scale, it's already been the most effective. TV measurement, some people have argued that in terms of adoption for OTT advertising, that measurement is the long pull. This is the bottleneck. It may very well be, because there hasn't been a good way to measure what television is doing to your offline sales, for instance.
There are a whole bunch of things that we have been doing to take your frequent buyer card and all the data-driven ways to track back to individual ads that are shown so that we can complete the cycle. You may say, well, digital isn't going to do super well in the themes that are not sold online. I think exactly the opposite. It's going to do well on all of it, but especially on the themes that are sold offline because the attribution and measurement is so bad in the status quo, there's an easier way to do it. Our go-to-market, Brian's going to talk a lot more about this. I want to just expose a little bit more visibility than what we have in the past. Our go-to-market is to full court press on the MVPDs. We have, we've had amazing adoption.
They've been a huge part of our success to date. Then there's a bunch of digital channels, especially those that are not in the skinny bundles. Whether that's a Hulu or whether that's Discovery Channel or whether that's any of the channels that aren't in the skinny bundle, we're having great success and great dialogue with those. Then I'll say that the last group of people, those are the Disneys and the Comcast. Those are the very big media companies, our dialogue with them has been unbelievable, I think we have great relationships built there. In some cases, we're already doing large amounts of spend with them. In terms of the massive adoption and us being the partner that I think we will be for them long term, the best days are ahead of us, especially in that last group.
We are full court press on all three of those at the same time, where the most amount of adoption is coming in the first two groups, as you would expect, in the early innings, which is where we are. Our long-term target is adjusted EBITDA at 40%. Our take rates will stay in the range where they've been. At times, we've brushed up against this already. I just want you to know that we're constantly looking for ways to reinvest that so that we can continue to grow as aggressively as we possibly can. We're not focused on maintaining EBITDA margins as high as they've been. We're willing to let that happen, because we're not going to invest in foolish ways, especially things that can mess up our culture. We think that would be the biggest mistake.
One of the common mindsets for people in our position is the view that money is free money. You don't get credit for the surplus EBITDA that you've created, why wouldn't you spend it and take a bet on something else? Even if you look at it as free, if that adds risk to your plan or adds risk to your culture, then it's not worth it. That's what we've repeatedly done. I just want you to know we continue to look for it. If there's ever a point where it drops lower, that's a good thing because we found things to prudently invest in. Just to highlight some of the things that I think make us successful and exciting and unique is that we're a self-service omni-channel platform. We're omni-channel, we're everywhere in the world.
There are lots of companies that are self-service that are just focused on individual markets. In this space, those won't exist long term. Either they're going to get global and be either a competitor or a partner to some part of the value chain for us, or they're going to go away, because I don't believe you have the luxury of just focusing on one geography. I believe that's especially true in China, where there's a lot of Chinese-only companies that have done okay as independents that are going to struggle as the game becomes more international for China. 95% of our spend is coming through MSAs and is predictable and, of course, recurring. Our tech is measurable. It's measurably better than our competition at decisioning. It's great that we can expose that measurability. We spend a lot of time talking about consumer surplus.
We shipped one product in Next Wave that actually reduced CPMs by 20%. We can justify our entire fee with just that one feature that we shipped. We just continue to add to the consumer surplus, and it's just proof that we're objective and we're aligning our interests with media buyers. $1 trillion is big. That's the reason why so many big companies are looking at it. We think we're better positioned to benefit from TV and international growth than any company in the world as it relates to advertising. We're delivering both revenue growth and profitability, which almost never happens in a company that especially has revenue acceleration. Of course, we've been profitable since 2013. If I were to summarize all that, we're a play on the whole Internet and not just a destination. I have two more slides.
The first one is, there are a bunch of people in this room who took bets on us when ad tech was the most hated category on Wall Street. I think we've together done really well, but I don't take you for granted. I just truly want to say thank you for believing in us at a time that others didn't, and for taking the time to listen, when in some cases others wouldn't. I take very seriously, we as a company take very seriously the concept of building trust. We've done that in China. We did that with Wall Street. I said when we went public, we are in the business of building trust. I said a year and a half ago at this meeting, we're in the business of building trust.
I recognize that our category has done a really bad job of doing that, and we think the only way to do it is to go public, do as good a job as you possibly can at making and keeping promises. I think we've done that well. I think we will continue to do that really well, but this is not possible without you believing in us, and I just want to thank you for being part of our success because you are part of our success. That's all I have for you. Thank you. CEO Jeff Green. Thanks, Jeff. We're going to take a 10-minute break now and reconvene with Founder and CTO, Dave Pickles. All right, let's get started. I know that coffee lines and bathroom lines are coming to a close, so I want to do the introductions today.
Most of the people you're going to hear from today are people that you, by you I mean investors, have not had much contact with. I deliberately, in order to help keep them focused, have asked Dave and others to spend less time interacting directly with Wall Street and investors and more focused on, of course, the task that they have at hand. There's not as much visibility into them, and I just want you to understand who they are. Today, I asked Chris if I could do introductions so that you could just get to know them very quickly. Dave Pickles is my co-founder. I would have never started this company without Dave. Before we'd started, I took him aside and said, "Hey, here's what I think we could do. Here's the business model, but I won't do it without you.
I need you to do it." He was game and bought into it, but there's no way that there's even a Trade Desk if Dave hadn't signed up for this. Dave is the best CTO I've ever met. Not just ever worked with, but I've just ever met or heard of. When I try to give people shorthand for why he's so impressive, I sometimes explain that his mother was a psychoanalyst, understood people really well, and his dad was a nuclear physicist, very into math and science. When you put those two things together, understand people really well, but your bias is really towards technology and science, that is the shortest way for me to introduce Dave Pickles. He understands people really well, but he also just loves geeking out on the technology.
This, I honestly believe, is a rare treat for investors to hear from Dave, who understands both of those dynamics, and he's going to come talk about our technology. Dave?
All right.
Thanks.
Can everybody hear me? I'll take that as a no. All right. There we go. All right, I'm Dave. Today I'm going to start with walking you through some brief history, because I think it'll help you understand not only more about what the Next Wave was and why that was really important, but also what we're going to release this year and why that all matters. I'm actually going to start by showing you a picture of our first UI. This is from 2011. It was created when we had about 10 employees. It had a back end to go along with it. It was really amazing, and it could handle maybe 100,000 queries per second, and we were super proud of it. It was incredibly basic, but it could do two things that turned out to be really important.
I touched on this last year, but just wanted to go over it quickly again, that bid factoring is one of the core differences that sets our platform apart, and we had this with that original UI. In most other DSPs, they use a decision tree to make all of their decisions. A decision tree is a data structure that says, I need to know everything about what I'm trying to bid on to be able to make any decisions about it. I need to have all the data in one place. I need to know that it's a 300 by 250, it's on a fitness site, it's in the Midwest, it's at 8:00 A.M. I need to know all those things at the same time. It creates a tremendous number of permutations to try to represent that.
You would need 24 line items to represent What we said was, actually this is a really pretty intuitive thing. I'm pretty sure whether or not 300 by 250 size of creative is working or not has nothing to do with whether or not it's in the Midwest. They really don't have to be considered together. Let's value all those vectors independently and then combine that knowledge to create a bid. Which it has a whole bunch of advantages for us, but it makes it so that we can be much more specific about what we want to buy and get a price that much more accurately follows the true value curve. I promise this is the most math you'll get in this presentation.
If the yellow line there is the perfect bid, that's the right price to pay for that impression, for that advertiser. It's something you'll never actually find, but it's the perfect bid. The line item systems do a really good job occasionally. That purple line goes and finds the perfect value because it found that magical combination of variables. They just didn't find it very often, and they weren't able to have enough RAM in a computer to store enough of the permutations to do it frequently. The factored bid never exactly matches the bid, but it gets very much close enough, and it does it every single time. I can follow that curve, and I can approximate a true value curve, and it just produces a lot of yield. Every time I underbid, I miss an opportunity and I miss scale.
Every time I overbid, I waste money, and I make the campaign perform worse. This is one of the core things that we had from the very beginning. It's really how we made our mark. We could do one other thing. Our UI had two features. The other feature was you could get a fantastic report. We worked a lot with the really early and shaky Hadoop stuff to get reports put out that allowed us to report on all of those individual variables that we were bidding on. No one else could do that because a tree, if you report on a tree, you look at node 5,000 in the tree, and I don't really know what that means.
If you look at a list of here's all the sites and how performance varies by site, it works for your brain, it works for performance, and we were able to put that report into Excel and hand it to our customers, and they loved us. We won a tremendous amount of business with an Excel report. That was our foundation, and this is just a random ship list from 2013 that I showed. We started adding things to it. We built our company to be a highly agile, highly iterative engineering team, and we started adding features. Some of the features were temporary. They didn't work out. We took them out. Most of them are really important things that we still use, like lookalike models for selecting audiences.
The ability to specify how bids should decay as data gets older, which turns out to be really important. Just all kinds of things like that. We added it in, we added in, we added it in. Pretty soon, we looked at the user experience that we were offering, which was this UI that we built, plus this huge group of features, and we didn't like the user experience we were offering to our customers. We started over. This was in 2013, six years ago. We started from scratch, and we shipped an entirely new UI, which is a theme that we'll repeat later in the presentation. We just looked at how we want our customers to be interacting with the platform. We thought this would be a better way to do it. We rewrote the whole thing, and it was fairly basic, again, by those standards.
It had all the features that the old one did, and we started again, and we started adding stuff. We added a lot of stuff. We shipped 3,000 features between, I think that's 2013 and 2018. 3,000 features got shipped to that UI, and they were important things. These are just some simple examples. They also give you an idea of what the scale of a feature is. It's just a simple thing that you can do. It's a new capability. We're putting things in the customer's hands all the time that are useful to them. Pretty soon that UI looked like this, which was pretty cluttered, honestly. It got to the point where it was really, really dense. This is what we showed you last time. It's now what we call Trade Desk Classic.
We decided to do it again, to start over and start from scratch. This is a piece of paper that I drew the design on for what I thought would be a better way to work with all this data. The core of this thought was, and you'll hear this again, too, I want people doing what people should be doing and machines doing what machines should be doing. If a person is doing a calc in an Excel file, that's something that probably should have been done for them. I want people making rich hypotheses. I want people dreaming and thinking about creatives and all the things that they're wonderful at.
I thought that by changing the UI so I could put the right kind of data in front of them on one screen, I could keep them from having to click in. It would be less of a drilling around, looking for data experience and more of like it's right there for you straight away. We wanted to use the screen real estate really efficiently, and that became Megagon. A Megagon is a little bit of a weird name for a product. My working title was 3.0, and nobody liked that, we had to name it something. A Megagon is a million-sided polygon, if you zoom out at all, it looks like a circle. It's something very, very complex that looks and is simple in appearance.
That sort of fit the theme of what we were going after. We had this great new UX. We had a team working on this for a long time, and during that same period of time, we also had an R&D team that had been really working magic with lookalike modeling. They realized that for years we had been lookalike modeling to say, like, "This is your core audience. What are some other audiences that have a lot of your people in them so we can go buy against that and get some good results? How do we buy third-party data effectively?" We realized we could use that same technique to also say, what geographies have a lot of your people in them, and what sites have a lot of your people on them?
What time of day has a lot of your people in it? We could do all of those different vectors. That's the technology that underpins Planner. Planner is a When you say Planner, you think of planning, but it's also an amazing tool for just campaign creation and optimization that's built around, there's this huge sky and all this opportunity. Let's run a model, and let's create a starting point that's smaller than the whole sky. I noticed that a lot of campaigns, you might be trying to spend $100,000, but you're targeting $10 million worth of inventory. Even though you're doing some good targeting, ultimately, you had some amount of randomness in what you were buying. There was this big pool that you were willing to buy from, and it was way bigger.
There were some obvious ways that you could say there's parts of that pool that are going to perform worse. Let's rein that all in and create a great starting point. When you use Planner to create a campaign, you can tell it who your target audience is, what kind of goal you're after. You can give it this information, and it'll come back to you with a bunch of options about here's a bunch of ways to spend budget, like what you said you wanted to spend or more, and you can take the pieces, and you can assemble it and go create a campaign that is going to perform really well day one. It'll continue to get better after that, but from the beginning, it's going to be pretty dialed in.
It still leaves enough room, so it's not so dialed in that there's no room to learn about this individual campaign and to fine-tune. There's a lot of balancing that went into that. It's an awesome tool. During that same time, we finally figured out after a lot of trying, how to get data scientists and engineers to really work together and to work together really efficiently and to have data scientists with PhDs and a very theoretical mindset to think in pragmatic ways and think about shipping software, which is really, really uncommon in data scientists. A lot of times, the output of a data science team is a research paper, and then an engineering team tries to figure out what the research means and go build something.
We got everybody to stretch and to form this real beast of a product development team that was just pushing out tons and tons and tons of really great AI-driven features, and that became Koa. Equally amazing, our marketing team took two years of three completely divergent work streams of huge amount of resources, and they summarized it perfectly in two words and put together just a fantastic piece of marketing. Our hard work has never looked so good. Jeff always accuses me of forgetting to paint the car. He's right. We're engineers. We build stuff and we ship it, and we know it's great, so we just put it in customers' hands, and they're going to love it, right?
This really did a great service to it, and it drove that adoption, which is good for our customers. It's pushed everyone ahead. I just wanted to give Susan and team a shout-out in front of everybody. They did an amazing job. Just like our first brand new UIs that we built from scratch, it's not an end, it's a beginning. This is a new foundation. We're going to continue to push hard and iterate on top of it. I'll shift gears now and I'll talk about the upcoming investments. Everybody's always asking me, what are we investing in? We put together a subset of what we're going to invest in. There's actually a lot. It's a really diverse platform.
There's a lot of work streams, but these are some of the really important things, I'll make sure I don't stand in front of them, that we're going after this year. My first theme was improve the number of campaigns that are executing optimally. I look at everything that's executing on the system. Some things are really fine-tuned and really effective, and some things have sort of just been loaded and forgotten. A lot of what we're doing with Next Wave is trying to address that.
We've carved out a huge amount of our resources in product and dev to work with our trading excellence team, which is a team of people that are just focused on what would be the right way to do trading, to build features that will make it so it's really efficient to go do all of the things that they should be doing. I always say, I want people doing the things people should be doing and machines doing the things machines should be doing. This allows us to do that much more efficiently. We also want the trader to be able to transition from any stage of media buying to any other stage of media buying. That's why I said Planner. Yeah, it's a Planner, but it's so much more. It's a toolset.
If you're later on, if you're in execution and something's going wrong, the AI's not doing what you want or whatever, you need to transition back and go use the AI tools and make that a seamless process. Go use the Planner again, go replan, push it back over. We're just trying to make it so that everybody can be hyper effective. Same thing, increase the number of campaigns executing optimally. This is the graph of the adoption of Megagon. I just wanted to show it visually because Jeff put the numbers up, this shows just the march that has been adoption. There's a couple of stairsteps in there where things moved over, basically, customers try it, they love it, the campaigns execute more optimally. It gets better performance.
We've been able to use the performance as a very strong hook to get people to move over because it just speaks for itself. The thing just works better. A lot of the adoption that is left is the customers with very complex setups, they're the ones that have the most disruption to shift gears and move things over, we're working with all of them to move over. I don't know of any cases where anyone doesn't want to move over. It's just a matter of getting the execution done, our product team's working on that. Another thing we're going to do is invest heavily in Koa. Koa is a type of a tree, I thought about it getting taller. That was where the slide came from. We've actually established an AI lab inside of the company.
Not every engineering team understands the AI toolset. It's really a specialty inside of engineering, in addition to data science. What we're trying to do is make it so that when we look at every problem that we're trying to solve as a company, we evaluate the off-the-shelf AI tools and the things that are available first as part of that. AI is not right for everything, but when it is, we want to take advantage of it. The hard part for us is that we usually can't use the off-the-shelf AI. We have to go soup it up to be able to operate it at very large scale. We're very good at breaking tools. 10 million QPS kind of breaks anything.
We're going to continue to invest in making sure in that person and machine balance, that the machine is an increasingly capable partner to the person in taking its share of the workload and doing a great job with it. Identity Alliance, we've talked about a lot too. It's what I consider and we consider to be an ideal approach for building a marketplace. We obviously have an in-house product with AdBrain for doing cross-device correlation, but we also have wonderful products in the platform from Drawbridge, Tapad, from Oracle, from LiveRamp, that each have their own strengths. As a platform, since we want to service everything, we have to support everything. Just because we have a product in-house doesn't mean we can't perfectly support everybody else.
We have to make it so that if you want to use Tapad, it is just as good or better on The Trade Desk than it is anywhere else. We have made sure of that. We realized we were in a really good position as a result of that, because we had all of these datasets that were all good at something or all good at lots of things, great at something specific, great at multiple things. If you overlaid them, if you could find a way to create a rule set that helps you pick which one to use, then you could use them all at the same time on a campaign. More than that, we could reassure our partners that we were serious about this marketplace thing and that we weren't just trying to shift everything to AdBrain.
We created a rule set in cooperation with all of these other vendors. We all said, "Here is how you should choose between the products. Here's the rules, and we run the rules." That just splits the spend accordingly, and you get better reach and better performance because of this. Just some visual sets of what that looks like. Instead of just using one graph, you're getting the power of all these graphs. The greater reach on your best data is almost the best thing you can do as a trader. You've got a very small, very specific strategy that works unbelievably well. The rest of the campaign is all these things that don't work as well. They work good too, but you try to blend it out.
You try to push as much money as you can into the best techniques, making those bigger makes the whole campaign work really, really well. This has been great for us. I think we can do this with a lot of products. This is the same position we're always in as a company. We have all these in-house efforts. We're building all of these tools that have corollaries outside, that have whole companies that do them. We need to support that company. We need to support ours. The additional benefit of this model is that it is going to continue to drive innovation. If you think that somebody has figured out the perfect way to do probabilistic cross-device correlation, you'd be wrong, right? There's a lot of different ways to do it. There's a lot of innovation that's occurring in the space.
By pushing each other around this rule set, like this is the right way to do it. We're all trying to improve our dataset to get some more market share. That makes us all get better, which makes us compete much more effectively with the walled gardens. It's a really great complete package and a good strategy for us on lots of types of data. Unified ID. This is also going to be fairly technical. To understand this one, this is a drawing that shows what happens when you start loading a webpage. You clear your cookies, you go to any webpage that is a publisher or an advertiser. This process is going to begin, it's because of the way cookies work.
With cookies, your ability to just put an identifier on someone, which is just a number, call Jeff number 1,234, I can only do that on my website, on my domain. I can't share that data with anyone else unless I send them the data that says Jeff 1,234, They send me back the data that says, "Oh, I've seen Jeff. He's ABC," We make a mapping of those things. This is one of the unfortunate things that we deal with as an industry with the design, we've had to go out and create all of these cross-connects. If you've ever loaded a webpage recently, you might notice that the status bar sits there and shows lots of activity. Because this is occurring behind the scenes until you're fully mapped and even thereafter to make sure everything is still in sync.
That's pretty gross. What's worse is when the mappings don't get done, Every time you go through a generation with a piece of data, You collect data, you push it somewhere else, you take that data, you push it somewhere else. You take your first-party data, you load it to a DMP, Your DMP sends it to your DSP, then you transact with an SSP. You lose reach every time you go through that because there's a drop-off in who has a match with whom. It's all these overlapping Venn diagrams. It's also just not optimal, it's what we've had to deal with. The only reason we had to deal with this is because the ad tech community were using cookie footprint as a competitive advantage.
Whoever was bigger wanted to have better reach and use that to go in, and we did that, and everybody did that, and it was really effective. At the point we're at now, we think it's better for us to try to help the sector grow than it is to try to continue to win on this. We have lots of other ways to win. We've said, what if we let everybody use our ID generator? Hopefully that'll look something more like this. What we first did is we went out to the supply side, and we said, "Hey, supply side, you guys can use our ID as your own, and then at least when you match with us, it'll be perfect." We went into the headers.
The very first thing that happens when you start transacting RTB, we establish that sync, and then they pass that ID to everyone else in the ecosystem. Instead of doing that whole out-of-band thing, we're actually using the flow of data that just naturally exists in RTB to propagate an ID. It's a very natural way to just get the ID pushed out to everyone. There's lots of numbers that we can push out on how effective that was because even just between us and the SSPs, we got the match so much better that we were able to have more reach on our best data. We also told those SSPs, the supply side, that they could share our ID with other buyers. This is where it really comes together and where the last benefit comes in.
If we also let other DSPs use our ID, it's all on the supply side. We don't need that syncing process anymore, and we can take it down. One of the headwinds that we face as a sector is that user experience can continue to get worse. As we add more companies and as we add more cookie mapping, there's delays, and there's all the page loading. There's a lot of sharing of data that I know everyone's very concerned about that comes along with those pixel syncs. Whereas ours is just an ID that's flowing through the pipes, when you do the cookie mapping, you can also learn many things about what's happening on somebody else's website. It's not ideal in lots of ways. It's definitely not ideal for consumers. We think we can help out a lot with that.
I actually often wonder what the carbon footprint of cookie mapping is, because it's this technical heavy lifting operation that really only serves to work around the weirdness of cookies, which were invented in 1994. Effectively, what we're trying to do is make the web space work more like in-app, because on the app space on your phone, it doesn't work like any of that. There's an identifier for your device, and you have controls on your phone to say whether or not you want it to be passed along, and it's the same for everybody. In-app works correctly. We're trying to just move the mobile web world and then the desktop web world in that direction. CTV is obviously a huge focus. We're really on a roll with CTV, and we intend to keep that going.
CTV has some pretty key differences between it and display or mobile. We can do a lot of our thing. Part of the value is just we do our thing, but now we do it in CTV. Just a couple of things to call out. The integrations can be technically different. There can be all kinds of complexities about the timing of when an auction occurs and what has to happen with the creatives. The creatives have to have better specs because they have to be high quality. There's all these things that we're figuring out with this sort of highly fragmented world of TV supply. We're trying to move everyone towards standards as best we can because we think that's best for everybody, but there's a lot of work in just getting all the integrations done. Another key difference is that it's not cookie driven.
It's not person to person or a lot of the things that people talk about with the display world. Many people experience the device, and oftentimes they experience it together. The data science is pretty different when you're trying to say what happened as a result of viewing these ads, and it's just a whole area for innovation. There is a lot of signals. Nobody ever tried before because in linear you got no data, or the best data you got was panels or things like that. Now there is tremendously rich data about what's happening that we can turn into better reporting and better effectiveness for our customers. The last thing is TV buyers are used to working towards different types of goals. You didn't go on to a linear TV buy and talk about a cost per acquisition.
You talk about reach and frequency. We have this entirely different set of goals that we're going after. We're retuning our forecasting to understand it. There's a lot to getting this right, to making it work correctly for the two different kinds of buyers. There's digital buyers that are kind of moving into TV, and then we also talk a lot about TV buyers and making a tool that's going to work for them in the way they think. A lot going on there. In China, I think it's a pretty fair assessment to say that China kind of has its own separate internet, and it definitely has its own separate market for everything that we work with.
While the core technology is the same, the bidder is the same, we've had to go reassemble an entire portfolio of inventory, of data, of quality verification tools. There's this whole portfolio of partnerships that we've needed to develop that we have, including new hosting infrastructure. Just getting a fast network connection, getting an ad from one side of China to the other is actually a very challenging thing to do. We've had to iterate a few times on who we're working with for hosting. We actually settled with Alibaba Cloud as a provider there.
We've done a lot of work to become even more cloud agnostic or environment agnostic than we were before because now we're operating in data centers we own, data centers we lease, Amazon, Alibaba, which is a good position for us to be in as an engineering team because being able to operate in all these different environments makes you have a lot of discipline in how you deploy your software. It makes you able to go faster and do better and be more efficient, which we have become. Then maybe even more important than all of that, we've built great teams in Hong Kong and Shanghai. We've been lucky to hire some amazing engineers in China to help us create a product that feels The Trade Desk and feels familiar. Like I said, it's not a totally disjointed experience if you're used to using our platform.
It's the same basic thing, but that it also speaks China and addresses the local market in the right way. This is mostly a reminder. This is something we keep investing in, but we have a lot of really successful customers on top of our APIs. You might know them as a DSP or something else, but they have taken our core platform and they've extended it. We built our whole platform in layers so you could consume it in pieces. You could use what you need from it and then build your own special sauce on top of it. We're super committed to the success of our API customers. As we go through all these UI iterations, we need to bring them all along with us. We're putting a lot of work in there.
We're also just trying to always make it easier to consume so that more customers can take advantage of it because it's super powerful and you can build lots of stuff on top of it. We started with that so that the very sophisticated customer can really do their thing and really flourish as a business with their own proprietary advantage and their own whole thing. We want to make it so that more and more people can take advantage of having their own custom The Trade Desk and extending it in ways that work for them. I'll close with some data we're really proud of. Jeff shared some of this too. We took two years and 40 years of our engineering resources to build Megagon, which is a huge investment for something that I drew on a piece of paper.
We really believed in it. As we've grown the engineering team, we've actually become slightly more efficient per engineer, which is a pretty big feat. We did that including coming into compliance with SOX, and going through all the training cycle and increasing the rate at which we're hiring. We've really got engineering management figured out in a way that I didn't even think we could do. I always told Jeff, the bigger the team gets, the slower we're going to go. It's just how it works. It's unavoidable. There's more network effect. There's more meetings that have to occur. The way we've been able to organize has made it so we can create eight Megagons this year.
It's a challenge for engineering leadership to keep our heads around everything that's happening, and we're spending all of our time just trying to make sure that these very fast teams that we've put together are building in the right direction and creating the right things, and we're super excited to see what they produce. Thank you.
All right. I didn't think of the overhead of introducing everybody. Next up. First of all, you can see why it's so much fun to work with Dave. He's just an amazing business partner. I'm also excited to introduce you to Brian Stempeck. Brian came on in the first 10 employees. We've worked together almost a decade. In fact, this week he'll celebrate his 40th birthday. We've experienced his entire 30s together. He came from Bain. He knew very little of our space before he got here. He's learned it as fast as anybody that I've ever met. Perhaps the best compliment that I can give Brian is that he is, aside from learning this super esoteric space and becoming very meta so that he can explain all of this, I think he's the best closer that I've ever met.
He's been on so many huge deals, and even today when we go into those meetings, he's probably the only person in the company that I find myself wanting to defer to because he'll help us get to the point and get to the punchline better than I would do. It's such a pleasure to work with him, and so without further ado, Brian Stempeck.
Cool. Thanks, Jeff. I thought we'd start with a quick quiz. Can someone explain Unified ID to us quickly? Just kidding. You don't have to explain. You don't have to explain Unified ID. It's hard. It takes a minute. I'm going to talk today about our customer relationships and then a bit of focus on why we think we're winning in connected TV. First off, as Chief Strategy Officer, a bit about what I do. I sit across our sales teams, data partnerships, inventory partnerships, which includes TV, and then we have a technical account management team. Ultimately, the role of my teams and myself is to figure out what are the big brands and agencies, what do they want to do next on our platform? How do we build the product and partnerships to accomplish that?
We work very closely with Dave and the full product team. A bit about our customers. Our split today is 60% large agencies, 40% mid-size agencies, and some of the enterprise API clients that Dave referenced. That hasn't really changed too much. When we did this investor day a year and a half ago, that split was 58%-42%, essentially the same kind of breakdown that we saw. We have a lot of clients who are building on top of us using our bidders, and they fall into kind of that second category. Our relationship with our agency clients continues to get deeper and deeper, this is a snapshot of a large agency we signed in 2011. You can see over time how the number of logins to the platform grew, and there's two big things that are happening here.
The first was when we launched the DSP and sold it to them in 2011, we were new. They were still testing us. They were still getting used to our platform, and they were using other platforms. Part of this growth is us winning share inside the agencies. What was also happening is by 2012, 2013, the agency was using our platform. They were buying with us at scale, but their teams were still small. Within the agencies themselves, they've been hiring more people to focus on programmatic. They've been training more people. Just overall, programmatic is powering more of the media plan to the point where you see in 2018, when you've got 1,200 people at an agency all logging into our platform on an everyday basis.
We expect that number to keep going up as we get deeper and deeper into each one of the agencies. Part of the reason that's happening is that we have this ongoing shift from the way people buy media. Historically in digital, they've gone direct to a publisher. They go to Hearst or Condé Nast or The New York Times, and they cut a direct insertion order. That's the bar in the light blue there at the top. 2016, that was 56% of digital spend, whereas 44% is the dark blue in programmatic. This is changing over time where it's essentially flip-flopped in the past couple of years where now the majority of spend is going into programmatic. The source of this data, we employ a firm called Advertiser Perceptions.
What this company does is they survey about 500 marketers, 500 agencies, and they ask them questions like, "In the next year, where are you going to spend your budget?" We get responses like this that help indicate, well, how are marketers thinking about this? Are they still bullish on programmatic? Are they putting more money into programmatic pipes? That's something that we continue to see that they're doing. As a result, the agencies have to follow suit because there's a lot of people at the agencies who are the ones managing insertion orders, they're managing ad servers, and they're the ones starting to learn how to use a DSP instead. Something that our business team overall thinks a lot about is expanding our core customer base.
The primary person in an agency who uses The Trade Desk or uses any DSP is a trader. We don't stop there. Most traders at this point inside an agency, they know who we are, they know our platform, they've been trained on it. They're typically trained maybe on a couple of DSPs, but they're experts. We're going deeper and deeper to say, well, how do we talk to the media planner? How do we talk to the investment lead? What about the TV buyer at the agency? As we get into connected TV, and at some agencies, you actually have a digital team and the linear team combining forces. How do we make sure we're talking to the right people?
As we hire new people on our teams, it may have been the case a few years back where one account manager at The Trade Desk might have called on 10 agencies and supported them. As we work with more and more people at the agency, our team is calling on fewer agencies. You as an account manager, you might be calling on two or three agencies instead of 10 you did before because we're going so much deeper and getting upstream on investment decisions. What sometimes happens as well is that we talk to the VP of Media, and that's typically the person on the brand side. There was a story about us recently where one of our clients was talking about our strategy, and I thought he explained it well.
He said, The Trade Desk, in many cases, we have a triangular relationship where the agency is at the table, the brand is at the table, and so is The Trade Desk. That's so the brand can have influence over the strategy and programmatic, then get more insights out of it. They can work with the agency on what they want to do. We're all in there together, where we're seeing the VP of Media alongside the agency team, all working together on programmatic campaigns. A big part of our success has also been driven by our global presence. This is a differentiator for us against a lot of our competition.
If I think about some of the large brands we work with, as they are growing globally and they're expanding into Asia, and they're saying, "Okay, we want to tap into the 300 million people who are in Indonesia, one of the fastest-growing middle classes on the planet." Many of our competitors simply don't have offices in that region, certainly not in Indonesia. We have teams on the ground to actually help train the agency and work with the brand to go after that market. Because while there's a lot of things that are the same, there are others that are pretty different. Our partnerships team will go into Indonesia and say, we have lots of global publishers, lots of global data sets. Who are the regional publishers that we need to bring into the platform? Who are the regional data sets?
Building capabilities inside our platform to support them, whether it's in South Asia, whether it's in China, whether it's in Latin America, that's something we do all over the world. We're constantly adding new supply. The way our partnerships team thinks about it is we're kind of stocking the shelves of the store. There's lots of things that agencies and brands want to do that are unique to different countries and markets, and that's a big place that we focus on. One trend that we get asked about a lot is the trend of in-housing, where you hear a lot of talk about, well, will advertisers take programmatic in-house? Will they actually get rid of the agency? Or will they take media overall in-house? If you just go by the headlines, I'm not sure that you get the full story.
You'll see things like, oh, two-thirds of brands are going in-house, or this story in Adweek, 86% of brands are taking programmatic in-house. If you read the fine print, who that's sponsored by, Accenture Interactive. They have a vested interest in making this feel like something you should do, because then as your consulting firm, they can advise you on why that's important. Bear that in mind with a little bit of a grain of salt. I want to talk about kind of the overall trend that we're seeing. Here's what we think in-house really means, and there's really kind of three flavors to it. The first is what I'd say is the student. If you go to an advertiser, they have a couple people internally, the VP of media, they're the internal media expert.
There's all the brand managers or the category managers or the GMs, and they're not expected to be experts on media. Instead, what they do is they go to this central person, they say, "Hey, you're the expert on media. Tell me how I should be informing my agency. Give me your kind of point of view on things." That's still a pretty common model with many advertisers. This person, they want to be up to speed on programmatic, but they're not going to go in-house, meaning they're not going to be hands-on-keyboard. They're not going to log into a DSP. They're not going to run campaigns. They simply can't. They're too short-staffed. They're the one responsible for, hey, what's happening with Snapchat? How much should we invest in Instagram? What should we be doing with digital out-of-home? They're thinking much more broadly than just programmatic.
The middle category, center of excellence, we're seeing more of this, where an advertiser might have five, 10, 15 people, again, who are experts in media and are spending more time in programmatic. We typically see these people engaging more with us as a DSP alongside their agency. They might log into the platform, but they're rarely running campaigns. They're doing that to pull analytics. They're doing that to pull insights. In some cases, they're doing that to say, "Hey, I want to check the homework of the DSP or check the homework of the agency, and I want to look for transparency. I want to make sure I know in that supply chain who's being paid what." Again, they're not in-house in terms of being hands on keyboard. They do want to be closer to us and other DSPs, but they're not truly in-house.
The final version is what I call what most of us think about as true in-housing. A brand that has a larger internal team that looks more like agency media buyers. They're hands-on-keyboard, they're using a DSP, they're running campaigns themselves, and they're not using the agency for that. That's the minority of what we see. Typically, when you do see it's with more acquisition-driven clients. Some insurance companies do this. Netflix is a famous example. They have a large in-house team. They don't use an agency. If you live and die in customer acquisition, especially online, you have a big search team internally or a big social team internally, it might make sense for you to have programmatic in-house. Again, we don't see that as much as the other two models. This is something we actually surveyed marketers about.
We asked them, "When you buy programmatic, how do you buy it?" The vast majority in the light blue here say, "Well, I use my agency to buy programmatic." The purple, they say, "Well, I use my agency trading desk," just the specialized part of the agency that focuses on programmatic. Then 29% say, "Well, we use our internal media team." When we first saw these results, we were like, wow, 29%, that feels high. We asked some follow-up questions, and we said, "Well, what does your internal media team do?" They said, "Well, they advise the agency on how we should be buying programmatic." The vast majority are actually using an agency, even though they might be talking like they're not. They are getting closer to programmatic. They're thinking more about it. It's powering more of their media plan.
From the data we've seen, we think about 10% of advertisers are truly going in-house. Lots are talking about it. Lots are thinking about what their role is. But in terms of truly hands-on-keyboard replacing the agency, we don't see that as a wholesale trend right now. The other point to note here is that sometimes, again, more in the media, in-housing feels binary. Either you're going to take this in-house or you're not. What we find as we talk to advertisers and agencies is that there's actually a bunch of different kind of chapters in a book when you think about, well, what do you want to get out of in-housing? One example would be, I mentioned with transparency. Well, if you want transparency in the supply chain, that doesn't mean you need to hire 20 or 30 people to go run a bunch of campaigns.
We can pass you log-level data. The agency can pass that data. If you want clarity in the supply chain, that's something you can get either way, and we're seeing more advertisers who log into the platform. A technology integration. We're seeing more marketers now saying, "Well, I actually want to link my sales data into my programmatic platform. When I optimize a campaign, I'd rather optimize to sales lift than I would to clicks or viewability or something else." That doesn't mean that they need to own the DSP contract or be in-house, but it doesn't mean that they need to get closer to us. Part of how our business team talks to our clients about this is to say, what is the goal of the advertiser here?
What does in-house mean to them, and how do we accomplish that without them having to scale up a team to do this themselves? We support a pretty wide range of engagement models from the left to right, kind of most common to least common. On the far left-hand side here, this is when the agency has a contract with The Trade Desk and the agency manages the campaigns. That's the most common model. Second most common, the agency has a contract with us, they manage campaigns, and they occasionally give logins to the brand. They give some level of access to the advertiser, also increasingly common. That middle path, we see kind of what I call almost the agency of the future. It can be consulting firms. Sometimes we see programmatic specialists, just programmatic-first agencies, and they're running campaigns. Again, they have the contract with us.
They're running it. Less common, the brand has a contract with us, and then they just give the agency access to run it on their behalf. We do see some brands who do that. The least common is the brand owns a contract with us, and then they say, "No, we don't actually have an agency. We're doing this in-house ourselves." We support all of these models today, but the vast majority of the business we run is on the left-hand side. As we get deeper with relationships with our clients, we're staffing against that. Dave touched on this when he talked about APIs and the things people are building on top of our platform. During some of the changes that Jeff mentioned with Alinea, we staff up some new teams. One was technical account managers.
The reason we have technical account managers is that the top 3 boxes here, so sales, account manager, and trader, that's always been our core team working with our clients. We started to add technical account managers, solutions architects involving our partnerships team, and they're helping build custom things. This is more like how Salesforce operates. Once you have Salesforce installed at your company, of course, there's custom things that you want to do that might be different from someone else. Salesforce does that themselves, or they have independent firms who will do that for you, where you can build out your own custom instance of Salesforce. As programmatic grows and takes over more of the media plan, that's something we're seeing a lot more of.
We have this technical account management team and the solutions team to say, "Okay, let's build in your sales data. Let's build a custom algorithm with you. What do you want to build on top of The Trade Desk?" Of course, that makes us stickier with our clients long term, along with driving better results for them. A few examples of that. One would be workflow automation. If you're an agency today, if you think about your day-to-day job and you're a media buyer, you might take the creative that you're going to run for a campaign, you load it into an ad server, you put in the flight dates, you put in the budget, all these details, you have to log into the DSP and do that all over again.
You're doing a lot of duplicative work that's kind of manual entry type thing. It doesn't have to be done twice. We've worked with agencies to say, "Well, let's link your ad server to us, so just do that once. Cut down the amount of time that you need to do." That's a custom build that we can do. We have other clients where they've gotten pretty sophisticated about how they build their own algorithms. They're buying media from The New York Times, and they're saying, "Well, I can buy that through 3 different SSPs. Which one is best? Where do I see the fewest price floors?
Where do I have the best ability to buy that inventory cheaply or get the most fill or get the types of ads that I want to reach?" They're building algorithms on top of our platform to optimize that and say, "I only want to buy The New York Times through OpenX or through Google or through different SSPs." These types of products are places that we're getting a lot closer to our clients. Overall, one trend we're seeing is that most of our clients historically have been using the goals on the left. These are pretty typical digital metrics. So whether it's a gross ratings point, you're measuring viewability of your media, you're measuring click-through rate, how on target is your media or the overall reach that you got. That's where the world primarily lives today in digital.
What's possible is that we can do all the things on the right. If you want to measure, you're selling an ice cream brand, and you want to measure sales lift of a new product driven by advertising, we have companies with rewards card data from grocery stores. We can measure the lift in sales. You're operating a retail chain. You want to see if, did advertising get more people to go to the store? We can measure that lift and use that as the objective for the campaign, and use all the AI and machine learning to say, "Let's hit that goal," instead of hitting a click-through rate goal. Whether it's brand awareness or media mix. We're encouraging our clients and offering products to tie business results to advertising much more closely than they have been in the past.
These things on the left, these are all valid goals. Of course, it's important to know that your ads are viewable. That's your media quality goal. When it comes to business results, the thing that a CMO cares about, that the CEO cares about, is did you drive sales? Did the ad campaign drive sales lift? We think, for the first time, we're giving them a very clear answer on whether that's happening or not. I'm going to talk a bit about our story in connected TV and why we think it's resonating. The core piece of this pitch, as we go to agencies and brands to talk about connected TV, is that without it, you're missing out on a huge swath of your audience. There's a large group of people that you're no longer reaching if you're only buying broadcast TV.
Sometimes the analogy we use is that in the 1970s, there was three broadcast networks. There was CBS, ABC, and NBC. If you were a TV buyer then, of course you had to buy all three. If you left CBS off of your plan, you'd be missing out on a huge part of America. We think that if you're not buying connected TV today, it's at about that scale. You're missing out on an audience that's at the scale of leaving off a big three network in the 1970s. That's a part where we have to educate marketers and agencies a bit. Jeff slightly touched on this, that streaming now is mainstream. It wasn't a few years ago, but at this point, with 75% of U.S. households that have an OTT device, we're not at the early adopter phase anymore. We're at late mainstream.
You can't actually buy a TV anymore that's not a smart TV. It's going to be internet-connected. It's going to have apps on it that allow you to stream content. Of course, the story about cable's decline. I think what's most interesting here is there's a lot of talk about cord cutting and cord shaving. In the fourth quarter of last year, the decline in cable subscriptions was down 4%. Two years ago, it was down 2%. Last year, it was down 3%, then it was down 4%. The pace of that decline is accelerating. There will come a point where that gets even larger, where cable starts shaving or cutting down 5%, down 10%, and that's where I think you start to see more and more dollars go into streaming, even faster than they already are today.
It's also important to note here that our message to clients is not connected TV versus linear TV. You have to buy both. We work with major brands who need to reach huge audiences. We can't go to them and say, "Just buy connected TV. Don't buy broadcast anymore." There's a huge number of people who are still on broadcast who do watch linear. We're telling them to buy both and how to buy it smartly. We're helping them use data to understand, here's the audience that you're reaching on broadcast, here's how you reach people you're not on connected TV. The usage continues to go up, so we now see one in five hours on TV is in connected or streaming. At that pace, by 2021, you're looking at two in five total hours of consumption.
Some markets out of the U.S. are already well ahead of that. Our reach is huge in this space. We have 80 million households that we can reach and 120 million connected TV devices. We do that because we're aggregating a very fragmented world. Whether you're looking at an OTT app like Tubi or Pluto TV, kind of smaller OTT native apps, or you're thinking about skinny bundles like Sling TV or DirecTV Now, or networks like Discovery where you can go straight to their app, we aggregate all of that. We don't really care how the consumer gets to that content. Our goal is simply to amass as much reach as we can across all those apps, across all those devices, to make it easier for the marketer to find people across all of that.
Sometimes when I'm presenting internally, you guys are sitting here for a lot of slides today, I'll say to the internal teams, if there's one slide I want you to pay attention to and remember throughout the entire day, this is the slide. As we got into TV a couple of years ago, we realized pretty quickly that you need to speak the language of TV buyers. The language of TV buyers is all about Nielsen and GRPs. That's how TV has been transacted for decades. How many gross ratings points did you get on a demo that you were trying to hit? 18 to 34-year-old males or females or whatever it was. We actually partnered with Nielsen, we did studies to say, "Nielsen, can you tell us, based on your panel, how many incremental GRPs we're getting?
What are we getting above and beyond what the brand already got from their broadcast buy?" We ran a bunch of campaigns, connected TV. We passed the data to Nielsen. We had them run it against their panel. What they told us was, all right, well, there's a bunch of people you only reached in linear. There's some who saw linear and connected TV together. There were a bunch of people who were only watching connected TV. In fact, across a bunch of studies that we ran with Nielsen, 41% of the households that we showed ads to had not seen any broadcast ads. Zero. This was the sales slide, right? This is the story to marketers and agencies to help them understand the scale of this transformation.
If you're a CMO and you run this study, and the answer is 5% or 10%, you don't really care. It's a rounding error. You can live with missing out on 5%. That's probably how it was four or five years ago. But when it's 40% of the people that you're trying to reach, you're not able to reach with broadcast TV ads anymore, they're seeing zero of your ads. That becomes a huge moment for you to say, "I need to move budgets to a different channel. I need to be spending connected in a way that I was not before." Having the validation of a company like Nielsen to understand we're speaking the same language, we're scoring this apples to apples the same way you did before.
This has been a huge reason for the fact, not just for Trade Desk, but for the entire category as a whole, why more and more spend from a media perspective is going into this space. Another question we often get is, well, I get the consumers are there, connected TV is more expensive. Is it worth it? Part of the appeal of broadcast TV is massive scale, massive reach, but it also can be cheaper. People have been negotiating for years. They've been buying at the upfronts. They're buying at scale. They get lower prices. Easy example that I'll walk through, you're buying broadcast at a $10 CPM. We'll talk about you're buying connected TV at a $20 CPM. The thing to keep in mind is that when you buy TV, for the most part, you're buying nationally.
When you try to actually explain that to a digital media buyer, it blows their mind because they're used to targeting at a zip code level or with lat-long coordinates or even at a city level. The fact that you can't target beyond nationally in broadcast is just so foreign to a digital buyer. When you go to optimize a campaign in TV, you can optimize on national delivery, on day parts, and maybe the shows. That's kind of it. You don't have a lot of control over what happens. Whereas in connected TV, it's just like everything else we do. You can choose which households see ads and which don't. You can do fine-grained geotargeting, pick different audiences you want to reach. You can frequency cap. You have much more control than you do with a national TV buy.
The question remains, is it worth it to pay double for connected TV? Some other studies we've done to look into this. Again, this is some Nielsen studies we did from our connected TV buy. One result we had, we came in at double the on-target percentage. On-target percentage means, okay, you're trying to reach 18 to 34-year-old females. 30% of the time you were hitting them. The other 70% was waste. We were doing that twice as well as broadcast TV. Again, not much of a surprise because we can just say, "Well, let's only target households that we believe are 18 to 34-year-old females," whereas if you're broadcasting nationally, you can't do that. We outperform there. The more eye-opening stat for a lot of our clients was on frequency.
We can control frequency at a household level or at a device level. We know we serve three ads to this Roku box, and that's all the brand wanted to serve. Let's stop serving ads. A national TV network or broadcaster does not have that ability. There's one household that's watching TV 10 hours a day, and the household next door is watching 30 minutes. The household watching 10 hours a day is going to see a lot more ads, and there's no way to turn off that ad load. You're serving that nationally. The efficiency you get from frequency easily pays for itself with a higher CPM. To kind of walk through that, here's how we think about it. Imagine a neighborhood. You've got 10 different households, and you're an automotive company, and you're trying to reach anybody in market for a car this year.
Those blue circles here, those are auto intenders. The gray ones are not a fit for you. If you're buying linear TV, let's imagine you're spending at a $10 CPM. For $1, you can get about 100 ads per household, right? That's probably like three months worth of ads, ad per day. You're hitting everybody in this neighborhood. You're spreading it pretty equally. Looks pretty attractive. You're getting mass reach. You're getting a lot of scale. It's pretty cheap. You're wasting a bunch of it. There's a bunch of households who aren't going to buy your car this year. They already bought a car. They're not interested in your brand. They've never been interested in your brand. It's not a good use of your resource. Worse, it's not actually $1.
It's only $1 per household if you could control it per household. With linear, you can't. Because of that problem of this house or the house with a bunch of bubbles next to it is watching 10 hours of TV a day, you showed 100 ads to them, they're never going to buy your car. You have no control over that. That frequency efficiency play is huge for marketers to understand, given limited budgets on marketing, where am I best going to spend that? There's a lot of waste built into existing models. Overall, even with a higher CPM, same number of ads, doing the same kind of math, you spend a lot less by targeting into the audience you want to reach and controlling frequency than you do by hitting a broad mass audience where you have a lot less control.
Couple of just quick case studies. The first is, we often have clients comparing us to other partners in their media plan. One would be user-generated content platform, large video play on the internet. You can probably guess who it is, but it's considered to be very cost-efficient. What we found was we were running connected TV, and we were outperforming them by 2x. They were looking at cost per completed view. That was their metric. We came in at half of what this other digital player was doing with better video completion rates. The primary reason we won this budget in the first place was because of brand safety concerns. A lot of marketers are scared off by user-generated content. They want to know what they're buying. What we're buying in TV is premium. This is network inventory. This is cable inventory.
They know what they're buying. We give them measurement and optimization on a bunch of different things. Whether that's looking at the total reach they get, the viewability, the video completion rates, or really important is reporting across screens. Something a lot of marketers are wrapping their heads around right now is, okay, I served three video ads on the tablet. Five on the smartphone, 10 on the OTT device, six on the desktop. There's waste built into that. Maybe I didn't need to have that household or that person that many times, but those systems are in silos. We've talked about Adbrain today and our device graph and the Identity Alliance. It's all aimed at solving that problem. How do you figure out the right number of ads to show people across different devices?
All these metrics let you get much more fine grain with the way you buy TV than the way you could do it historically. With our device graph, we can also do something pretty interesting, which is tie TV to business results. Of course, when you're watching your Apple TV and you're streaming something and you see an ad, there's no way to actually click that ad. There's no way to measure, well, did someone click that and then buy that product, either online or offline. With our device graph, we can measure that. We can measure, we showed you an ad on connected TV, you went to your desktop computer and bought something. We can show lift in that. We take data from companies like Oracle and Datalogix or Factual, which is a company that has store visits.
We can say, well, your TV campaign actually drove more people to the store. Your TV campaign drove more people to buy things at the grocery store. Instead of just looking for reach and frequency, and again, proxies for business results, today, we have clients using this and saying, "Tie connected TV to actual business results." This is a retailer. What they did with us, they ran connected TV ads. They used Factual as their location partner to measure after you saw an ad, were you more likely to go to the brick-and-mortar store? What they measured with us was, on average, they could drive someone to the store for $17. That was their cost per acquisition with TV. Really for the first time, as a marketer, you can understand beyond anecdotally, did my TV advertising work? Did I get people to the store?
Did they buy something? You can measure it at a very fine grain and start to optimize the campaign and direct it based on that. I can't say enough how much of a change that is in advertising. Marketers haven't had this toolset before to say, does TV advertising, the biggest part of my budget, does it work or not? We're giving them those tools. This is one of my favorite stories. An automotive brand, one of the smaller brands, they don't have the budget to compete with large spenders. They can't go sponsor Monday Night Football like Toyota or like GM, like a bigger automotive brand. This is a smaller brand. What they did with us is they built a custom audience. They said, "We know who has owned our car before. We know who's been to our brand website. We've got first-party data.
We've got some third-party data about who's in market. We've built an audience of 5 million people we think are most likely to buy our car." All we're going to do is instead of buying 18-34 female or 18-34 male or large demographic blocks like that, our new source of truth is this audience. It's people in market for our brand, in market for our car. We're going to focus on this, and we're now going to do all of our TV buying, measuring reach and frequency against that audience. That's the shift we're seeing with a lot of marketers is going away from age, gender, kind of more simplistic methods, and getting into the actual audiences they build and believe in and measuring reach and frequency there. Last couple things we're seeing.
For the first time, we're seeing programmatic play a role in the upfronts. Every spring, largely in New York, there's the upfronts where the agencies and the brands sit down with the large networks, and billions of dollars are transacted on a guaranteed basis for the upcoming TV broadcast year. For the first time, we didn't see this during last year's upfronts, we're seeing buyers, agencies, and brands go into those conversations with networks saying, "I'm willing to commit money to you. I'm willing to commit the $100 million that I committed last year to your network. I want a portion of that that I can buy programmatically, that I can put on The Trade Desk or other DSPs and buy with my audience data." It's a win-win.
The networks are also on board with this because from their point of view, they're still getting what they really want, which is the committed dollars. This isn't going straight into scatter market. This isn't going to bought on the fly. The commitments remain. The way that you buy the commitment is changing, and it's changing from the broadcaster saying, "I'm going to do this. I'll control GRPs. I'll decide what you get," into, "I'll let you buy this on a DSP. I'll let you buy the streaming inventory, layer your tools, your audience on top of that." That's a massive change that we're just starting to see happen at the 2019 upfronts. All of this has been a huge part of the reason for a lot of the growth that Jeff pointed out. We've seen growth in inventory where networks and MVPDs are coming to the table.
We've seen growth from buyers where they're coming to the table, spend up over 9X. The story is working, the trend keeps happening, we're super excited about what's going to happen next. Thank you so much for your time.
Just wanted to make sure I could promise you a break after the next one. We have one more. I just want to introduce Susan Vobejda. I mentioned that Brian was one of the first 10 employees, so we've been at this for a long time. Susan joined us a couple of years ago, when I was first introduced to Susan, I was unsure if somebody could come from fashion into tech. Of course, when you look closer to resume and you see that she spent a lot of time at Yahoo, she spent a lot of time at Bloomberg, of course, she knew tech. I was also pretty honored, honestly, to be meeting the CMO of Tory Burch.
Tory Burch had been a brand that I hadn't heard of very many years before that, then it had become a very premium brand in fashion. One of the things that intrigued me a lot about Tory Burch, and you may know this about high-end fashion, that it's not uncommon for the majority of high-end fashion brands' purchases to be done inside of China. In terms of an up-and-coming brand, there wasn't any brand that I was more impressed by what they had done in China than Tory Burch. As I got to know Susan better, I just became so fascinated.
Sometimes when you interview people, it takes a long time. After hours and hours you're like, "This is the right person." Susan was one where, in the first 20 minutes, I was like, "This is too good to be true." I just kept asking questions to just make sure that there was something that I wasn't seeing. She's one of the best CMOs in the world. She's the favorite one I've ever worked with closely, and she's also one of the most incredible human beings. I'm so excited to introduce you to Susan Vobejda.
Thanks, Jeff. Hi, everyone. That's me, Susan Vobejda. I'm so happy to be here today. I hope you're enjoying Investor Day 2019 with The Trade Desk. We've spoken a lot today about the fact that marketing has really changed in the past few years and is experiencing disruption. What I'm going to do is spend a little time talking to you about what it's like to be a CMO today in the midst of all of this disruption. What is the mindset of a CMO in the digital age like? As we all know from our jobs, including you on the investing side, there is so much change going on for all of us. Actually, there are few functions that are seeing the outright upending that marketing has been seeing in the past few years. Marketing used to be a simple and neat affair.
In fact, this is what I signed up for when I went into marketing in 1990. You made a few commercials a year as a CMO. You ran them through mass media channels. You reached almost everyone through three television networks. You had a lot of time for three-martini lunches and golfing with your agency. This was a snap, relatively. As a CMO, the heyday of 20th century marketing was made famous by the TV show "Mad Men." As a CMO, you could see those advertising executives constantly entertaining their clients. The role of CMO and the role of marketing has evolved to become one of the most diverse functions that you can imagine. Digital has changed everything about marketing. I've seen it up close. That's what's driving the success of The Trade Desk today.
The rise of the internet has fragmented audiences, as Brian just spoke about. Traditional media no longer yields the power to reach audiences efficiently and capture hearts and minds as it once did. Marketing is being disrupted. The roadkill of this disruption is the martini-drinking, golfing CMO. I'm particularly sad about the golfing. The CMO role today is diverse and unrecognizable. The ground that you're covering is vast, from strategy to creative, from analytics to IT. The responsibilities of a CMO can now encompass revenue and growth responsibilities, managing the biggest IT budgets in the company, the mandate to be creative and tell brand stories that capture the imagination of customers. Also a digital and analytical fluency to rival anyone in a C-suite, with maybe the exception of the CFO, Paul. The list of themes is long.
Because no CMO is actually a master of all of these things, we need assistance. To be successful, the modern CMO needs experts and solutions. That's where The Trade Desk comes in. Marketers and their agencies leverage platforms like The Trade Desk to efficiently engage customers, drive revenue, and move their businesses forward. Why has this change happened over the past decade? At the center of this change, the force that's driving all of this is the new digitally fluent customer, and that's all of us, you, me, our friends, family, all of us. Marketing has changed because customers have changed. Life in the past 10 years has changed exponentially fast for us, and so have our expectations. I want to show you a slide to really illustrate the force that digital has had in the past 10 years or so.
This is two images in the same location, and this location is Vatican Square. It's a similar event. In 2005, at the top, you see an event for Pope John Paul II. In 2013, you see the announcement of Pope Francis. This really illustrates that how we as human beings are experiencing life has changed dramatically in a short time. Global smartphone penetration has surged since the launch of the iPhone in January 2007. Today, the smartphone is an extension of each of us and is our way to connect with the world. Currently, about 38% of the global population has a smartphone, with 71% in North America, 68% in Western Europe, and 38% in Asia-Pacific owning smartphones.
The importance of mobile and digital to consumers cannot be overstated. As we've discussed today, mobile ad spend is approaching 50% of our business at The Trade Desk. This chart, this next chart shows how smartphones have impacted our time. The headline here is that we're all time-starved. Why are we time-starved? Because we're spending a lot more time in recent years on our phones. This slide shows time spent by U.S. adults on digital media each day. In 2011, you can see that U.S. adults spent 3.7 hours a day consuming digital media and less than an hour a day on their phones. Fast-forward to 2017, it's going to be similar when the 2018 numbers come out. We spend 5.9 hours, basically 6 hours a day, consuming digital media, with 3.3 hours spent on our phones.
We've actually lost about two and a half hours of time every day to our smartphones since 2011. If you've been feeling a little bit busy in recent years, this is why. We're swamped because of our phones. For marketers, it's harder and harder to actually get your attention. This is known as the attention economy in marketing. When we are able to get your attention, we usually have it for about six seconds. Where is this two hours of time for each of us coming from every day? You might think, "Gosh, my relationship's been suffering a little bit," or, "I'm not getting as much sleep as I used to." One place it's coming from is definitely print media. Worldwide print usage declined from 44 minutes a day in 2011 to 16 minutes a day in 2018.
Another thing we're doing much less of, as Brian has discussed, is we're watching much less of linear television. Linear television consumption dropped 30 minutes a day since 2015, and this trend is even more pronounced among Gen Z audiences. Linear TV viewers 18-24 years old dropped 15% since 2015. By the end of 2019, almost 30 million U.S. households will have cut the cord, and by 2022, we expect that number to be 55 million U.S. households cutting the cord. When you add that to the 40 million U.S. households by 2022 that are expected to never subscribe to cable, that means that 95 million U.S. households won't have cable by 2022. As we're cutting the cord and spending more time with our digital devices and not sleeping as much, what are we doing? We are consuming more digital content than ever before.
This slide shows what happens during 60 seconds on the internet according to The Trade Desk. Our data scientists took a snapshot of our bid stream at 6:00 A.M. on a Thursday morning in January. This graphic shows internet traffic and the corresponding ad impressions pattern in this 60-second period across the entire open internet. You can see that on the left-hand side, your left, some big sites like weather.com, CNN, Spotify, they get quite a bit of traffic, but pretty quickly, audiences scatter. The ad impressions come from 100,000 sites and total approximately 600 million avails. This really illustrates both the amazing reach and scale of The Trade Desk, but also the fragmentation of digital users. Contrary to what you've been reading, as we've been saying today, we're not just on two sites. Notably, 48% of these avails are from mobile devices.
As smartphones have proliferated, so have apps. App usage accounts for over 80% of mobile time spent globally. For The Trade Desk, mobile in-app revenue is one of our fastest-growing segments, up almost 100% year-over-year. Digital video consumption is another way we're spending time. U.S. adults in 2019 will spend an average of an hour and 32 minutes watching digital video. This is up 30 minutes a day from 2015, and of this, an average of 25 minutes is connected television. These numbers are actually higher in many countries around the world. If you're a brand marketer, you absolutely have to be thinking about how to create video and video advertising to connect with your digital audience. On The Trade Desk platform, all digital video, including desktop and mobile, accounted for about one-third of our business in 2018. What does this mean for CMOs?
The heavy consumption of digital media across very fragmented channels has changed marketing forever. Brands and their agencies now need to use data and technology to get in front of the right consumers as they move across media channels and across devices. We, as brands, need to do this cost efficiently and at scale. This is what programmatic media does at its core, and why programmatic is the way that a large chunk of digital media transactions now happen globally. According to Magna Global, in 2019, approximately 50% of digital display and a rising portion of mobile advertising will be RTB programmatic. To bring to life the importance of programmatic to CMOs, I want to continue with the automotive example.
In a given week, there are approximately 4 million automotive new car buyers in the U.S., and these active buyers are in the market for about 108 days, according to Cox Automotive Customer Journey Study. If this was 10 years ago and you were a new car shopper, you would see an ad or two on TV, you would scan a newspaper ad and look at some pricing, and then you'd go to dealer after dealer to figure out what car you really wanted to buy. Now, almost all U.S. car purchases are researched online, and it's not just a little research that we're doing. The average new car buyer spends six hours researching their car purchase online before headed to a dealer. For marketers in the automotive category, if you aren't in front of active car buyers online, you are not going to receive a dealer visit.
I was at a CMO conference in November. The CMO of a leading car company came barreling up to me. He didn't sprint up to me, but it wasn't like a slow walk. He definitely sought me out to tell me that The Trade Desk has completely changed his role. He could now reach the active car shopping audience much more efficiently and measure the effectiveness of his ads. This CMO and his agency use The Trade Desk for display, mobile video, and CTV advertising. The video format is of course key to automotive brands because it best displays the beauty, the design of the automobile, and the driving experience, and makes it more emotional. This CMO and his media agency also use our offline measurement tools to track dealer visits directly attributable to his programmatic campaigns on The Trade Desk.
He would bring this data to meetings and be able to show how many people he was driving into what dealers. Then his management team would then look at the conversion rates and compare the performance of each dealer. This just really changed the dynamics in the management and in the C-suite for him. He really brought some data to the table, which was key. Leveraging the power of data and technology for CMOs and their agency partners is critical today. Data that helps us understand a customer's journey is at the heart of marketing. Data is the currency of the modern CMO, just like it was in this auto example, and for this CMO. We have to connect with our audience at the moment of truth with the right message that resonates with him or her personally.
This is at the core of what programmatic enables and what we do here at The Trade Desk. In fact, as Brian talked a bit about, brands and CMOs are now working more closely than ever with The Trade Desk and their agencies. As programmatic becomes a larger part of their marketing budgets, brands want to sign an MSA with us, their DSP partner. They have meetings with the DSP partner, The Trade Desk, and the agency. When you hear us talking about new MSAs that have been signed, note that 90% of those MSAs also include an agency partner. It's really not, as Brian talked about, a company that's just planning on executing programmatic on their own. Another thing that we're seeing within our relationship with brands and with agencies is a sharp rise in graduates of our programmatic certification program called The Trading Academy.
We now have 10,000 graduates, including agency traders, agency executives, and brand marketers. I know there's actually a few of you in here that are also graduates. Because of The Trade Desk's progress in the market and the desire of agencies and marketers to work more closely with us, we've made significant gains in the last year, not just in our business, but as a brand. In looking at our brand data, which I'm going to show you next, we've seen a sharp increase in awareness, reported usage, and future intent when our customers are surveyed than we did a year ago. We've actually had a breakout year as a brand, and our position relative to others in the category is very significantly changed. Here's a slide that shows that. This is Advertiser Perceptions.
Again, Brian talked about this company as an expert in the ad tech industry, and they do a tracking study that's used by almost every brand in our category. This is overall brand awareness, 2017 to 2018. You can see overall that The Trade Desk moved from number 6 to number 3 in one year. If you break that down among agencies and marketers separately, you can see our big jump in awareness among agencies. If you look at Google, Amazon, and The Trade Desk, the top 3 among agencies, there's a bit of space between the three of us and the rest of the pack in overall awareness. With marketers, we saw the same thing, really jumping to number 3 in awareness among marketers in the last year. That's just awareness.
Those are the awareness numbers, which just is the top of the funnel all the way on your right. If you go across the funnel, consideration, usage, future intention, you see that The Trade Desk is a solid number 3 across all. When you look at net promoter score, which is likelihood to recommend, it's actually a measure of customer satisfaction. You can see that we're tied for number 1 here, which is something we're really proud of because relationships and our partnerships with our clients is so important to us as a brand. As Jeff said, I've had a long and varied career in marketing, and I became a CMO in 2015, and I was the CMO of Tory Burch, as he mentioned.
While I was there, we undertook a pretty significant digital transformation of our marketing efforts. As I was experiencing that, one thing I realized is that the closer I was to the technology and the data sets that were available to me, the more effective I was as a CMO in developing strategies, creative concepts, and improving efficiency and ROI. So 15 months ago, I changed my career trajectory. Here's a little scrapbook for all of you of my time at The Trade Desk. I started working at The Trade Desk and sort of bet my future a bit on marketing and technology really dominating what's ahead. I met the force of nature that's Jeff Green, and I met the brilliant Dave Pickles, and all of the incredible team members of The Trade Desk.
When you bet something like your career, sometimes you may look back a little bit or kind of have second thoughts, and honestly, I never have. I am so thankful every day to be a part of the company, The Trade Desk, and also a part of the future of marketing. Thank you so much.
Susan Vobejda, our Chief Marketing Officer. We are going to take a 10-minute break and reconvene at about 4:15 for Tim Sims, Senior VP of Inventory Partnerships. Our next speaker, we've got two more and then we'll do Q&A. Our next speaker is Tim Sims. He's our SVP of Inventory Partnerships. One of the byproducts of having inventory become more and more fragmented, and as I mentioned in my presentation that there's this asymmetry of power. There's this power concentration happening on the buy side and fragmentation happening on the sell side. It often makes it so that we can carefully choose the people that we hire from the sell side, and especially those that we have interacting with the sell side, we're going to choose very carefully. In my view, we hired the best of the best from Tim Sims.
His last job was at OpenX, the company we were talking about before. He was there for many years during their ascent. Before that, he was at SpotRunner. One of the very first experiments in sort of ad tech for television. He brought all that experience to The Trade Desk to be on the buy side. He manages our partnerships as well as I think anybody could. He's one of the most valuable assets that we have here, Tim Sims.
All right. Thank you, Jeff. Afternoon, everybody. As Jeff mentioned, I'm Tim Sims. I'm the SVP of Inventory Partnerships. I've been with the company for about five years. Prior to that, I was on the sell side of our industry. In some ways, I've watched the programmatic industry mature for about a decade. Today we're going to spend the majority of the time going deeper on the supply side, and in particular, we're going to take a deep dive into connected television. This is a question we get asked a lot, especially in rooms like this, to just go a little bit deeper, one or two clicks deeper into the actual supply chain of connected television. We're going to spend the majority of the time today doing that and what that means for the future of television.
First, let's talk a little bit about what the heck Inventory Partnerships does and how that helps drive growth for the company. The role of the organization is relatively straightforward. We oversee global partnerships with everyone on the supply side. Another way to think about that is if you sell ads for a living, you're my team's purview. I also oversee the global technical account management team that Brian mentioned. You can think about TAM as the folks who help everyone plug into the platform, whether you're a partner on the inventory side, a data partner, or a client who's building something on top of our platform. If you're interacting in a technical way with our platform, TAM is the running point on that. Our team oversees partnerships with global marketplace operators like Google and Oath, Rubicon, Telaria, and others.
We're also responsible for our global media relationships with big media like NBC, Disney, RTL Group, Tencent, Alibaba, Spotify, Pandora, pretty much anyone on the supply side of our industry in the digital ecosystem. Over the years, our platform has expanded dramatically what we have access to. When I joined 5 years ago, we were largely a display video and mobile company. We've expanded into native audio and now connected television. In a sentence, this is really how I describe what my team does. When a customer logs into The Trade Desk, they should have access to every impression on every screen everywhere in the world. That's a super lofty goal, but that really is the North Star for my team, and that's what we think about every single day when we get up and come into work.
How do we make this statement come true for our clients on the platform? It's really important that we do that because as the world is evolving, it's becoming more and more connected. To some extent, inventory is the lifeblood of our platform. In order to live up to everything that Jeff talked about earlier and power data-driven decisioning at scale everywhere in the world, we have to have inventory to do that. When you take data and point it at the Internet, you need access to every screen everywhere in the world for a brand to have a cohesive conversation with a diverse global consumer that's changing the way that they're interacting with content everywhere in the world. How do we do that? We do that with a big team.
We've got over 50 team members in 14 offices around the world, which includes feet on the ground in markets like Shanghai, Tokyo, Hong Kong, Singapore, and of course, the United States and Europe. The core functions of our team, as I said before, are kind of managing existing partnerships. We have a lot of existing relationships with the marketplace operators that I mentioned before, but we also manage new business. Of course, business development is a really important part of what we do here, especially in new markets like Southeast Asia and China and Europe. I want to take a second and just talk about one of the themes that really drives that business development side of what inventory does, which is move fast and be first. This is a mantra that our team talks about.
We spend an enormous amount of time really thinking about speed to market. Dave's org that he spoke about earlier allows us to be really agile and bring things to our clients that they've never seen before, and in some cases, bring an opportunity to a partner that they hadn't even thought of yet. I'm going to walk through an example of what this looks like in practice, just so you can get a sense for what we're actually doing on a day-to-day basis. Joox Audio is a company that you've probably never heard of. They're the largest streaming service in Southeast Asia. You can think about them as the Spotify of Malaysia, Thailand, and Indonesia. They're also a division of Tencent. A member of our team identified this opportunity in Singapore and thought it would be something great that we could bring to our clients.
Through another team member in Hong Kong, we arranged a meeting with Tencent in Shenzhen and met with the Joox team. Prior to that meeting, Joox had never considered programmatic. They had never monetized an impression in a programmatic ecosystem ever before. We talked to them about our experience in the market. We talked about the success that our clients were having running programmatic campaigns across Southeast Asia. We introduced the idea of making their inventory available on our platform. They'd never done this before, so they didn't have a tech partner. They didn't know how to make their inventory available. We introduced them to an engineering-led SSP out of Taiwan called ucfunnel. ucfunnel did custom work for Joox Audio so that they could make their inventory available on our platform.
In August of 2018, we went live with Joox Audio everywhere in Southeast Asia, and today we're still the only DSP that has access to this inventory in the region. I don't expect that to always be the case. I expect other demand to come on for Joox Audio. With that mantra of move fast and be first, we were able to bring something to our clients that they'd never seen before and go to a partner and introduce a concept that they had not even yet considered. This is what we do on the inventory team every single day. Another theme that we think a lot about is how do we continue to expand the depth and breadth of what we have to offer on the platform? Oftentimes, I think about the team's job, in the context of a spreadsheet.
If you look across the top, that's every single channel we could possibly be in. If you look down the vertical axis, it's every single publisher that we could possibly work with in the world. Our team's job is to fill in all those cells and make sure that we've got a checkbox everywhere. Nowhere is the expansion of our platform more starkly evident than what we've seen in connected television. The headlines that we've shared on earnings calls, this has been referenced several times today, those are the headlines, right? 10X growth in 2017, 6X growth in 2018. I think what we're going to spend the majority of the time doing is going one click deeper onto what's actually driving this growth. What is the force that's pushing this forward?
Before we do that, I want to just maybe spend one minute on a history lesson of how we got to this point. If you rewind three years ago, maybe three and a half years ago, Jeff and Brian and myself, we'd go sit in meetings with content owners and streaming services, we'd talk to them about our company. We'd introduce the concept of programmatic, that meeting, on average, went something like this. "I don't know, guys. I'm sold out. I'm just trying to get my subscribers up. Programmatic? I don't know.
We'll see." If you fast-forward to three years later to today, all of those same companies that the three of us sat down with three years ago are some of our biggest partners in connected television executing programmatically today, that's because they recognize programmatic as a viable option for them to monetize their content. That shift, that 180-degree shift, is what we're going to dive deeper into today with a little bit of a guided tour of the CTV supply landscape. First and most importantly, let's define it. What do we mean when we say connected television? This is what we mean. We basically mean a TV that is internet-enabled. The means by which that television becomes internet-enabled doesn't really matter to us.
It can be an OTT device, it could be an OTT stick, it could be a gaming console, it could be the television itself. As long as that television is connected to the internet, we can deliver an ad into that environment. Interestingly, these devices also play a role in how we kind of understand the household from a targetability perspective. Through our acquisition of Adbrain that Dave talked about and through data science that we have internally, we're able to triangulate around identifiers and devices to understand audience and households so that we can more effectively target at that level. When we zoom out and look at that and what that means from a connected television perspective, we have access to 80 million households and 120 million connected television devices. That's really meaningful.
These numbers are pretty big, especially given that connected television is still kind of growing as a channel, we're still adopting it. We're all still in this adoption phase of CTV, that's a pretty meaningful number. What does that mean? What does that enable? What does a footprint like 80 million households actually do for us? One of the core drivers of spend growth on our platform in connected television is the ability for an agency to take an audience and apply that to TV. If you think about it from a brand's perspective, just take a CPG, take an auto, take any brand that you can think of. These guys spend an enormous amount of time, effort, and money trying to understand the digital persona of their customer and their future customer.
Now, for one of the first times in the history of marketing, they can take all of that hard work and point it at the television screen. That opportunity is unprecedented, and we're just at the beginning. CTV is still small. We're still in the first inning or whatever sports analogy we want to use. I sometimes joke that we're still tossing the ball around the outfield, right? The game hasn't even started. When I look at this, all I see is opportunity. Everybody in this room is voting with our viewing habits, and that's helping to dictate the future of television. We're all changing the future of television by the way that we're consuming content, and that's a really powerful force. One byproduct of all of us changing the way that we view content is that the supply chain starts to fragment.
Gone are the days of these three or four major networks. You do a couple upfront buys, and you're done. That world is over. In the connected television world, the sellers have a very siloed landscape, and they're somewhat beholden to their own geo footprint. If you think about this, when you go to a single streaming service or single device manufacturer, they only have access to the people who have subscribed to that service or actually purchased that device. That is the media footprint that you have to work with if you're a media buyer. That same fragmentation existed in digital years ago. Imagine a world maybe 10 or 15 years ago where you had to actually call up 50 websites to place an ad on the homepage. That's super inefficient.
One thing that programmatic does really well is it aggregates all of that fragmentation. Just to take a look at what this fragmentation looks like in a real-world example, I pulled some data for you guys. This is from The Trade Desk. This is CTV avails on our platform from a single inventory source. Again, if you look at this through the lens of a media buyer, if I'm going to go to one source, one streaming service, one device maker, one publisher, I only have access to the geo footprint that they have from their subscriber base. This is what that looks like on a map. This is avails over a seven-day period cut by zip code. That's what that fragmentation looks like from a single source. That's what you have access to. You're beholden to that geo footprint.
When we think about the reach, Brian talked about this earlier, we have an enormous amount of reach in our platform. You take that 80 million households, and then you couple that with the fact that we can partner with everybody in the supply chain. We can partner with streaming services, we can partner with the content owners, we can partner with the device manufacturers, we can partner with everyone. When you couple that with the reach that we have on the connected television device graph, by contrast, this is what it looks like on the platform. This is all inventory sources, CTV avails by zip code on The Trade Desk platform over a seven-day period. That's a powerful slide because essentially what we've done is we've aggregated the national footprint, and we've aggregated what effectively is a national buy on television.
Here's the beauty of this. If you want to buy national, great. Amazing. We can do that. If you want to buy the top 20 DMAs, great. Amazing. We can do that, too. If you want to buy the 42 zip codes where you have a dealership location, amazing. All of that is possible. What's more, if you want to overlay audience, manage frequency holistically, all of that is possible with basic targeting features in the platform. That really is the opportunity, and that's indicative of the reach that Brian talked about earlier. This is what it looks like on a map on The Trade Desk. That's, I guess, a good segue because when we think about reach, that's only half of the story, right? Can we reach a large audience in the United States? Yes.
The question I'm asked most often is, what exactly is the inventory? I've been asked this question by a few folks in this room because conceptually, all of this makes sense, right? I have an Apple TV. I get that people are cutting the cord. I read all the headlines. I understand that Samsung and LG are never going to make a television that's not a smart TV. I get all of that, but what exactly is the inventory? We're gonna spend the next five minutes or so talking a little bit more about that. Before we transition into that, let's just ground this in a concept, which is that the television is now just a screen with apps. This is just like your phone. There are apps on the television screen through which I access content.
If we think about it in that context, everything else starts to be a little bit easier to grasp. How are people watching connected television today? We talked about this on an earlier slide, but they're using these devices. They're using a smart TV, OTT devices, gaming consoles, so that they can access content from these guys. On earnings calls in the past, we've bucketed inventory roughly in this same way. We've talked about streaming services or kind of the OTT natives, I think as Brian said. We've talked about direct-to-consumer, and we've talked about the virtual MVPDs. This is how people are actually consuming content on these devices from these guys. These are the three primary inventory sources that we look at when we look at the landscape of connected television. Let's zoom in for a second and see what this actually looks like.
We talked a little bit about distribution fragmentation, also what's happening in parallel to this is content consolidation. Jeff talked about some of the M&A activity that's happened over the last few years. You've got Disney and Fox, Comcast and Sky, all of this other stuff that's happening. Meanwhile, in parallel to that, you've got this distribution fragmentation. You've got a direct-to-consumer app. You can think about that as like a Discovery GO app or an app that's produced by the content owner that's on the television screen. You can watch it there, but probably they did a deal to distribute content with a streaming service, probably they did something that looks a little bit like a carriage deal with a virtual MVPD, they may have done exactly that same type of deal with the OTT device.
The nice thing about this is that if you notice, this also aligns perfectly with the way that we describe the four primary buckets of inventory. These three guys, and to a lesser extent, the folks who operate the OTT devices, are all able to sell inventory to us. When you lay this kind of supply chain on its side, this is essentially what it looks like. You've got premium content owned by the content owners going through this first layer of distribution between devices, streaming services, and virtual MVPDs. All of us in this room are consuming that content on these applications. That's really what it looks like. If you think about it in this way, hit TV show distributed through a streaming service, viewed on an app on a smart TV, that's really the end-to-end flow of that content.
We partner with every single person in this chain. We partner with the people at the end of the chain, we partner with the devices, the streaming services, the virtual MVPDs, and the content owners, which makes it a really viable market for us because we can partner with everyone. The mechanics of this, of how we actually deliver the ad into television, look a lot like what we do in other channels, like display and video. These apps on the television call an ad server. That ad server calls an SSP to conduct an auction. The DSP then bids in that auction. SSP clears auction, passes values into the ad server, and then the ads serve. This construct is nearly identical to the way that it works in display and video and other channels that are inside of programmatic. It's not this kind of foreign concept.
There's a lot of work to do, there's a lot of partnerships and integrations that have to happen, but the basic mechanics are the same for how we interact with television and ultimately deliver the ad to that screen. Because once again, it's just a screen with apps. I want to circle back to this for one second because I want to make one point really clear here that I've been asked in the past. If we just zoom out and focus or kind of get rid of those other ones and focus on the devices. When we buy an ad from the content owner or the streaming service or the virtual MVPD, we buy that ad in a device-agnostic way. As a result, we're able to run ads on any device that enables that television to be connected to the internet.
That could be an Amazon Fire or Roku, it could be the TV itself. It doesn't matter who's in this purple box. When we buy the ad from these guys, it's run in a device-agnostic way. It's a really important concept to kind of separate the device manufacturers from the content owners and the streaming services and the virtual MVPDs. This also then informs a lot of how our strategy works when we think about how we approach big media companies. Going upstream is a concept we talk a lot about kind of on the client side, and Brian walked through a good example here about how we get closer to the media buyers and the media planners and often work closely with the brand as well.
When I talk about upstream, I mean the media companies, I mean Disney and NBC, because we want to be closer to the content owner, because ultimately they control the majority of the ad rights downstream from the content. Let me walk through an example of what this looks like in practice. I think Germany provides a really instructive example here because they've done a really good job of vertically integrating at some of the large media companies in Germany. This is RTL Group or Bertelsmann, but they also own two SSPs. They own Smartclip and they own SpotX. SpotX is based here in the U.S., and Smartclip is in Europe. They also own ad sales houses that handle a lot of the direct sales for their media. They also own television stations, and then of course, RTL Group and the parent, which is Bertelsmann.
The natural kind of way that we want to move from a strategic perspective is exactly that way. We want to start going upstream. We've had longstanding relationships with SpotX and Smartclip. We've been working with them for years. We also want to work with these guys. We want to work with the ad sales houses. We want to work with the corporate executives who are helping to shape the future of television. We want to be in that conversation as much as we can because when we get into that conversation, it turns out to be a really, really fruitful exercise. Here's what I mean by that.
As we move upstream and have more of these conversations inside of big media companies, the objectivity that we talk about on the demand side, this ability to kind of make an objective decision for an advertiser or a brand to buy this impression, not that one, that point cuts in both directions. When we sit down with a media company, the fact that we don't own a large user-generated content video site, the fact that we don't own a movie studio, the fact that we don't own a television network, these are good things when we get in these conversations because we're a really attractive partner.
If you couple that just with our basic business model, the transparency, the self-service, and just the way that we operate our business and the decisions that we've made to never own media, never build sell side tech, all of this makes us one of the best possible partners for these large media companies because they look at us simply as a way to connect to demand that wants to buy television this way now. That's the way that we communicate that message when we get into these conversations more and more. Inevitably, that conversation also turns to the guaranteed market. The upfronts are a very efficient model, and as Brian said, we're never in this conversation to say, "Forget about the upfronts. Forget about linear." We do go into these conversations talking about both.
Guaranteed inevitably comes up because that's how the majority of television is transacted. My team spends a lot of time thinking about how the programmatic guaranteed market is going to evolve, not just in television, but in all of digital. Whenever we start to think about this concept, I always come back to this notion, which is reimagining our fundamentals. Agencies and brands make commitments and guarantees to publishers. That's the backbone of the media industry. That is how everything works. Programmatic is just making that process a little bit better, or arguably a lot better. Platforms like The Trade Desk are in service to that transaction. That's the fundamental of the media business. Now, programmatic has introduced a lot of new concepts. We've introduced things like programmatic guaranteed and private marketplaces to this notion.
The way I think about all of this is that these are just order types that are a way for a brand or an agency to execute a media buy with a content owner. If you think about the collection of all of these different order types, and how guarantees evolve, then over time, it starts to look something that's not dissimilar from portfolio management. I have a bunch of different strategies of how I'm going to engage with that media company if I'm a brand or an agency, and it's our job at The Trade Desk to make sure that they're allocating those strategies appropriately and making sure that if PG's not working, we reallocate to the open market. If this isn't working, we reallocate here.
Our job is to help them make effective decisions, but all in service to that core fundamental, which is that agencies and brands make commitments and guarantees to publishers and content owners. Now looking ahead, my team has a lot of areas of focus, clearly the stuff that we talked about today. I'm going to highlight three other areas where my team is really focused on what's next. The first one is inventory growth in live television. About a year ago, I sat in a lot of meetings at CES, and I sat in meetings with broadcasters and networks, and one consistent theme that emerged out of that conversation was something like this, "Hey, I've got a whole bunch of live inventory, and I don't really know what to do with it.
Can you help?" Of course, Brian and myself and others who were in these meetings were like, "Yes, we'll help." Essentially what that is indicative of is that what's happening in streaming is that the content owners and the folks who own the broadcast rights to live events are having a very difficult time predicting how many people are going to end up on the live stream. There's not a ton of historical information here, right? We're only a couple of years into this thing, so how many people are going to stream Monday Night Football or a football game on Sunday, doesn't have a lot of historical data. What ends up happening is we have all these spikes in inventory, and what programmatic is really good at is helping buyers be opportunistic and take advantage of situations like this.
We built a lot of technology on our side to help manage these situations where you've got this spiky and lumpy delivery of inventory. As a result, we were able to run a fair amount of inventory and budget across live events that we've highlighted in earnings calls in the past. Things like National Football League football and college football, Major League Baseball and things like that. We see this as another area where we can continue to move fast and be first and really innovate around live inventory. The next is international growth. As I pointed out, we've got over 50 people on the team, but I've also got people in seven offices outside of the U.S. We talked a lot about fragmentation earlier.
Fragmentation is very present internationally because not only do you have geo fragmentation, but you also have language fragmentation, and you've got various content owners and providers in all of these markets. What we're very good at is normalizing that fragmentation so that brands and agencies can activate data everywhere in the world. When we look at areas of growth and opportunity for our team on inventory, we definitely are looking internationally. Nowhere is that more true than in China. My team's responsible for the deals that we make with the Tencents and Baidus of the world, the Youkus and the iQIYIs. All of these deals that we've gotten in place with inventory providers in China are part of my team's purview, and we're looking forward to what's next in this market.
I think one of the most fascinating things about China is just in the details of the numbers. My favorite stat about China is if you look at internet use per 100 people in 2000, it was 1.7. So 1.7 people in China had access to the internet in the year 2000. If you fast-forward to 2017, 54 people out of 100 had access to the internet. Roughly 700 million people got access to the internet in less than a generation in China. Said a different way, two U.S. got access to the internet in less than a generation. What happened as a result of that is that services like Tencent and Baidu and Alibaba were born from that.
What we're seeing today is that services like Toutiao and Douyin, which are the two core products for ByteDance, are able to go from zero users to nearly a billion users in just a couple of years. Not only is that indicative of what's available to us on the supply side, but it's also indicative of what the opportunity is for the brands as we start to launch in China. Just to really quickly recap some of the things that we talked about. Inventory is growing on CTV. My team is going to continue to move fast and be first and march ahead with these types of really strategic conversations with big media companies so that we can continue to aggregate all of this fragmented media landscape and bring that offering to our clients.
Just a really quick final thought before we wrap up. I wanted to give a little bit of context to why I care so much. Clearly I'm excited to talk about all of this stuff today. I gave this presentation to my kids last night. I wanted to run through it once and just make sure it worked. After I was done, I said, "What'd you guys think?" My six-year-old said, "You talked about TV a lot. What's that about?" I said, "Well, TV's changing a lot." I was like, "When I was your age, if I wanted to watch a TV show and that show wasn't on till Thursday, I had to wait until Thursday to watch that TV show." She was like, "Huh, that's weird." Walked away.
That was her takeaway from the entire presentation was, "Huh, that's weird." I think that actually tells the story of why I'm personally so excited to be in the role that I'm in and doing what I'm doing is that it's not going to be undone. My kids will never know a world that's not connected. We can't walk it back at this point. It's happening. We get to sit at the table and help shape the future of this new connected world and help enable our brands and our advertisers that are on our platform every day to activate data across this new connected world. That gets me excited every single day to come into work. Thank you so much.
All right. One more. Every time I listen to him, I get excited. He speaks on our all-hands. I leave every time. Often I have to follow him on our all-hands. We'll have 10 people speak. He'll be number 9, I'll be number 10. Every time I'm like, "Oh, I should change the order," because he's a tough act to follow. Paul has a tough assignment ahead. It was about 2 years before we went public that we made the decision that we needed to be what we were calling IPO-ready. One way that we were not even close to being IPO-ready was in our finance department. We didn't even have a CFO at the time. We'd had one previously, but we had a great controller. We had just a total vacancy in the CFO seat.
I think our finance team was less than 10 people. As you know, in a world of SOX compliance, having a finance team of less than 10 people virtually makes it impossible to be IPO-ready. Of course, we looked for somebody who could build up that finance function very quickly, had been through the process of being a publicly traded CFO. Had also assembled teams in the area. Being a publicly traded CFO in our area, had done it multiple times. Had also had a lot of expertise in audit himself to just make certain that we were doing the very best that we could. That included remediating material weakness, which we've been doing basically since we were public. Super excited that all of that's behind us. None of that would've been possible.
To be IPO-ready in less than two years, and then also go public with the track record that we have, wouldn't have been possible without the work of Paul Ross.
All right. Hello, everyone, and thank you all for coming today. As Jeff mentioned, I'm Paul Ross. I'm the CFO of The Trade Desk, coming up on five years now. I joined the company in 2014, and as Jeff mentioned, have been focused on building the accounting and finance team, financial controls and reporting, and of course, getting the company ready for life as a public company. Today, I'm going to go through our financial principles and give you a little insight as to how we think about these three key areas. This includes our growth, the investments we make, and how we think about managing capital. First, revenue growth. Revenue growth is our main metric, and Jeff mentions this on pretty much every earnings conference call.
He'll use phrases like, "Our focus is to grab share," or, "It's land grab time." In a programmatic market that is only 3% or maybe 4% penetrated, that we believe at end state will go to 100% programmatic, it's really economically irrational not to focus on growth because we're still early. Since 2011, when we had our first penny of share on our platform, we've grown faster than the programmatic industry and have gained market share. Our growth rate has consistently been at multiples of both the digital and programmatic industry growth rates, and we expect for that to continue. In fact, based on Magna Global's estimate for total RTB programmatic spend, The Trade Desk has gained market share every single year. The programmatic industry has been tracking to roughly a 32% growth CAGR through 2018, while ours has been about 80%.
We have consistently grown faster than the industry by at least 2X. The same is the story for digital according to IDC. We have continually grown at multiples of the digital universe every year. Cool. Customer count also continues to grow. We've averaged about 15% growth the last couple of years. When bringing on new customers, we look for quality over quantity, as we look for partners that can grow with us. Typically, these new customers are smaller independent agencies or digital solutions companies, who in turn bring more and more advertisers onto the platform. Through 2018, we were approaching nearly 50,000 advertisers. When customers come onto the platform, they stay. Our customer retention rate has remained above 95% for 20 quarters in a row. 2014 will likely remain our largest cohort ever.
This is the year that the largest global agencies in the world joined the platform, our cohort and the associated CAGRs for each year has remained very healthy. Nearly all of our revenue comes from MSAs, master service agreements, more importantly, over 95% of our business remains self-service. We are the tech layer only for the brand, not the service layer, this has not changed over the years, we expect it not to change. It's especially key for the operating leverage that we enjoy. Spend on the platform has been consistently diversified across many industries. The other category that you see is a compilation of many different categories and has been on top for two years in a row now. In 2018, health and fitness was the next largest at 15%, followed by technology and computing at 13%. On to how we think about investments.
As I just mentioned, revenue growth is our primary metric. The investments we've made in our people and technology are what's led that historical growth. Areas such as CTV, cross-device, Identity Alliance, and of course, international, are all areas where we invested early and are now seeing returns. We invest as fast as we can and would invest faster if the opportunities present themselves. Even though we want to invest as fast as we can, there are some limits. Finding the right people, the right cultural fit, and focusing on areas most likely to generate massive returns can sometimes slow the pace of our investments. As such, for EBITDA, we have maintained an operating plan in the high 20s as a percentage of revenue, which has been reflected in our EBITDA guidance consistently now for the last three years.
We generally expect it to remain in this range, we would invest more if we can when the right opportunities present themselves. Relative to our model, we have historically outperformed our operational plan. When that occurs, much of the incremental revenue above and beyond our expectations falls down into adjusted EBITDA, this is why we have ended the last three years with over 30% EBITDA margins, which illustrates the leverage inherent in our business model. While our longer-term EBITDA target remains unchanged at 40%, we are not yet managing the business to optimize to this 40%. If that was the only goal, we could actually be there now, we've instead opted to invest for what we see as many years of significant growth. Just a quick point on our take rate.
As Jeff mentioned earlier, we've said it consistently for a number of years now, we expect that long-term take rate to be in the range between 15% and 20%. It's ticked up a little bit over that range in 2018, we would expect that to move back down over time as more spend comes onto the platform and more volume discounts kick in. The first spend on our platform was in 2011, the first quarter of our profitability was in 2012. Again, a testament to the tech as well as the business model. We've generated real GAAP net income every year since 2013. Keep in mind, this is while we are investing as fast as we possibly can. Which leads us to how we manage our capital. First, a couple comments on working capital management.
It's true that in general, our payment terms are such that we pay before we collect. However, we've been focused on reducing the amount of days delta between our DSOs and DPOs, and you can see the positive downward trend in this chart. We utilize sequential liability terms wherever we can, which essentially means we don't pay until we get paid, and credit risk on an absolute basis is the lowest it's ever been. Secondarily, we've been more aggressive on collecting receivables and increasing our payment terms, particularly for newer vendors. My favorite slide. As a result, our cash balance continues to increase. We paid off the small amount of debt we had early in 2018 and ended the year with $207 million and no debt on our balance sheet.
Coupled with the $150 million available on our revolver, we have more than enough flexibility on our balance sheet to self-fund our growth, invest for the future, evaluate M&A opportunities, and provide ourselves the flexibility to weather any macroeconomic storms on the horizon. Finally, I'd like to close out my presentation with showing what The Trade Desk has achieved relative to other companies out there. This is how we stack up against other SaaS and internet companies in our world. We have posted revenue growth rates much higher than the fastest-growing companies in the SaaS universe. Our revenue growth just accelerated in 2018, which is almost unheard of for companies our size. Our profitability, even while investing as fast and as much as we possibly can, puts us in the same league as the largest mega-cap names in the space.
With that, I will turn it back over to Jeff. Thank you very much.
Thanks, Paul. We're going to take just roughly about a five-minute break, set up some seats here, have the management team come back up. Jeff's going to say just a few thoughts in terms of closing and then move into Q&A and then the reception. Thanks.
We have more questions than we have time, we'll get started pretty quickly. I just wanted to start this section by just saying thank you to everybody for being here today. Of course, thanks for your support as investors and analysts who spend a lot of time covering what we do. Just wanted to open it up for questions. In the future, I anticipate us having the entire C-team here. We don't have everybody represented, in part because two of the members of our C-team joined most recently, and our Chief Talent Officer literally, I've been emailing with her all day as we're hiring, as well as our Chief People Officer, who started on Monday. Then our Chief Legal Officer's in the front row. If you see her raise her hand, that's because I said something I wasn't supposed to.
Delighted that Vivian's here as well. Wanted to open it up for questions. Shyam? I'll let Chris figure out the order. Chris gets to decide where he goes.
Thanks, Jeff. Shyam Patil, Susquehanna. Thanks, Jeff, for doing the day. It's been very helpful. I had two questions. First one is, during your speech, Jeff, you talked about how you haven't even reached the tip of the iceberg from the DoubleClick ID change. I know it benefited you guys a little bit last year, but can you just talk about what you mean by this and just what the impact could be to Trade Desk over time? Then second question is, on Amazon, there's this belief in the industry that they have a tremendous data advantage over everyone else, and that that's going to help them with their demand-side platform and monetizing outside their walled garden. Can you just give your thoughts on that, maybe help dispel the myth? Thank you.
Yeah. I'm going to answer in reverse order because I think the second question has some bearing on the first. One of the thesis that we had at the very beginning of starting the company was that the most valuable data to any given brand is their own. This is maybe the strategic advantage that is a byproduct of our objectivity. Objectivity is sort of the head, but maybe that fact is even more of an advantage for us than the objectivity itself.
The reason for that is because when you have a huge amount of data, and I had this exact same conversation with Tencent in Hong Kong a week and a half ago, which is it's really difficult for them to walk in to a brand and say, "Trust me with your data," in part because of the rich data asset that they have. If you can couple those two things together, then it creates more asymmetry in their partnership. That creates way more risk in China, by the way. While Facebook, Amazon, and Google have to decide whether that's strategically worth it for them, you also have to add the legal risk that comes in with Baidu, Alibaba, and Tencent.
It's just something that they cannot do strategically at a minimum, but it is a huge advantage for us to then be able to win the trust of those brands. I do think Amazon has a tremendous amount of data, but the way that that typically works, and this is exactly the way I expect it to work with Tencent, it'll work the same, I expect, long term with Amazon, which is that they'll go to an advertiser and say, "If you want to use my data," and Tencent could argue that theirs is the very best in China, "then when you want to use my data, you have to buy it through me. When you want to use your data, you have to buy it through someone else," and we think we're going to be the most preferred someone else.
We think both can coexist, and Amazon will continue to do a good job of monetizing their own inventory for quite a long time. The first part of your question, remind me again.
DoubleClick ID.
Oh, yeah, the DoubleClick ID. The reason why I say we've only reached the tip of the iceberg is because you may feel like you're only beginning to understand the ramifications of it and the way all the IDs work. Marketers feel exactly the same way. They're still trying to understand all the ramifications. Google hasn't even extracted it completely yet, so it's only happened in places. They haven't even taken out the functionality in full the way that they will. Partly because they're giving marketers some runway to find alternatives. Because people tend to wait till the last minute, just in general, as well as the fact that people are still trying to understand what all of this means, I think we've only begun to see sort of the movement over that I think we will in the future.
I also believe that because I also said we're only seeing the tip of the iceberg in terms of what people are doing with data, that often data is sitting on the sidelines, and as a result, they're not really using the IDs to their full extent, meaning the ID is the way that you put your data to work. Given you're not putting your data to work, the loss of the ID is not as big a deal as it will be five years from now when I'm trying to put a lot of data to work and I couldn't possibly do it without the ID. Every day that goes by, the lack of that functionality becomes more important, not less, in terms of just the hole that it's leaving. Every day that goes by, it creates more and more of an opportunity for us.
My biggest anxiety on this is capturing the full extent of the opportunity fast enough. It's not have we overstated the opportunity, it's that are we up for it? There's just that much, that fast. I worry more about that side of it.
Thank you.
Sure. I'll let Chris figure out where to go.
Hi, guys, and again, thank you very much for doing this. Youssef Squali, SunTrust. Two questions for Jeff. Jeff, you said that the sum of the tech take rate is 25%-30% right now with the sell side and the exchange side maybe being a little too high. I think you're by 20% of that 25%-30%. Why do you think that is? How far do you think that goes down in aggregate? On the Amazon question, just a follow-up. I think you said that Amazon has a bigger objectivity problem with clients, and you gave the CPG problem. How do you see yourselves kind of helping them do that, and are they amenable or open to letting you help them?
Sure. All right, Amazon on the second question as a take rate on the first question. The numbers that we put up, which is the 20%-30% take rate, is where we think it will be at end state. I actually think when you look at all the taxes between there, the walled garden tax, if you will, and the independent internet are still, and I'm just guessing, 50%-ish. When you put ad serving costs, all the overhead of every part of the supply chain, it's still at 50%. I think at end state it's going to be 20%-30%. I think buy-side tech is going to be 15%-20% of that. We're currently on the high side.
Most of the trimming that has to be done is going to be done from the sell side, because as their business models have made that choice to focus on being pipes, the same way that I don't think the Nasdaq or the NYSE can charge 20% for being pipes. As markets scale, the intelligence moves to the edges, meaning that all the insight and the data asymmetry in financial markets lives with buyers and sellers. That exact same thing, I think, happens in the markets that we are participating in. Just as they're nascent, there's this ability to charge a higher tax, and it just takes time for the market to shine that light of price discovery and transparency on where fees are coming out, and that pushes those rates down. It's just a hypothesis that they'll be in that range.
You may say, "Why will they be even that high? I mean, if they're at 50, why do you think they'll be as high as 20-30?" It's a lot of triangulation on the lack of fungibility and the uniqueness in the asset that is advertising, where there's no one thing that's exactly the same as another that makes it so the volatility in our industry is so much higher than what it is in almost any other financial or commodities market that we think makes it so that the data differences in those groups make it so that those take rates will continue for a long time. I don't think there's a lot of controversy actually in our space about that. I do think that as that pressure continues to happen, actually, the appeal of the independent internet goes up.
I think we're only now at that tipping point where we can even make the claims that we are, which is that we should be the first stop on a media plan, walled garden should get the leftovers, non-data driven should get those leftovers. Take rate and Amazon. Remind me of the Amazon- Objectivity. Oh, their objectivity specifically. How you can help them. I think if you're Amazon, let me just, without overstating I'm just going to speak from the cheap seats, right? I'm not giving inside information. We have conversations with Amazon on a fairly regular basis, but I'm not going to disclose anything from those discussions. If I'm Amazon, I'm of course looking at trying to monetize my site first and foremost.
With the amount of traffic that we have, we are effectively the second-largest search engine in everywhere other than China. We have this huge opportunity that we need to monetize our site, and that's going to come first and foremost. As I'm also looking for ways to move the needle on my P&L, what I would be doing is how can I be a marketplace, and learn from what Play Store has done really well, which is arguably the greatest asset that came with Android. I think the same thing is true for the iPhone, for Apple's app marketplace. I think they tend to think in marketplace terms. How could we be a marketplace?
As they're thinking that way, I hope that we do a great job, as Tim did earlier today, of explaining that we are going to be a source of demand if they ever want to do that. Hopefully that characterizes it well. I don't know if you want to add anything or Brian?
I would just add to that. Facebook and Google have the objectivity problem of saying, if you go to Facebook or Google and say, "Design my media mix for me." They'd be like, "Spend it all on YouTube," right? I'm being facetious, but they bias towards their own properties. Amazon has that challenge. They have their own site. Amazon, I think, has a bigger challenge with marketers. At one point last year with our sales team, we looked at the top 100 advertisers on our platform, and something like 80 of them were competing directly with Amazon. Whether that's a tech company competing on cloud or a CPG company competing on baby wipes, or a grocery store retailer competing with Whole Foods. Google and Facebook don't generally have that problem, that their core business is in media and publishing.
Whereas if you're a CPG CMO thinking, "Am I funding my own demise by?" There is that conversation happening with a lot of big advertisers. The data is great. They do want to spend there at times, but they do so in a very wary way.
Hello. Here in the back. TV advertising question. Question number 1: can you help me reconcile? TV advertising traditionally is all about premium inventory, premium environment where brands are safe, and classic programmatic doesn't care about the environment, right? As long as you have the guy who meets 42 criteria, that's your guy. How do you reconcile that in going into TV where networks are really, really protective of their prime time inventory? If you come to him and say, "We should sell it programmatically," he will say to you, "This is the best asset I have. I use it to build relationships, I use it to upsell." How do those conversations happen? Whether it's scripted or sports, I would be really curious how you convince them to give up this leverage that they've had for dozens of years.
Brian, do you want to take the first stab at it, and then I'll just supplement?
Yeah. I think what Tim said about live is a great example, where if you were to pick one, if you're NBC and your flagship show is "This Is Us," you're totally right. They're not going to come to us and say, "Hey, Trade Desk, can you help us monetize 'This Is Us?'" They're doing that just fine. That's their tent or ABC monetizing the Oscars. They're great at doing that, and that's what they lead with. Where we always start with any channel, whether it's connected TV or if it was display 10 years ago, is with the places that aren't sold as well. In TV, the nice part about that is that it tends to be pretty premium versus where programmatic started in display, more longer tail sites.
The fact that we're only a couple of years in to connected TV, and we're talking about running ads on "remnant inventory" on the World Cup and the World Series and Major League Baseball and the NFL, that is premium inventory. Because of things like the unpredictability of, okay, the game went into overtime and we didn't sell that's a challenge for the network. They can either bonus more ads to the existing advertiser where they're giving away for free, but create a frequency problem, or they can turn to partners like us and say, "Can you help us monetize this better?" It can be giving them incremental new advertisers.
In many cases, we go to a network and they might say, "Hey, we're sold out." We say, "Well, at what price point?" They'd say, "Well, we're sold out at $15 CPMs." We'll say to them, "Okay, we'll set up the deal on our platform such that we only win demand above $15." When we do that, an automaker or CPG or whoever, they're willing to pay more for their own audience. We only bring the demand that's higher than that. That starts to get the wheel turning with the network saying, "Oh, wow, there's incremental new advertisers here. There's CPMs that are higher than maybe what direct sales got. This is a net positive to me overall." Your point is right.
They're not starting with "This Is Us." They dip their toe in the water with other types of channels, but they've seen good results, which makes them more comfortable doing more with us.
I'll just add a little color, which is that I do think you're exactly right, that programmatic was the rhetoric of those of us that were selling programmatic a decade ago was the show doesn't matter, basically. It's all about the audience. We would even say, because this is at least directionally correct, that the audience was just the show was just being used as proxy for audience. You would say, "Oh, it's Major League Baseball, and therefore the audience is probably male." Then you then extract all the things that you think comprise the average Major League Baseball fan. The pendulum swung too far the other direction, where we're like, we don't care about the content at all. We just care about the audience on the other side. Then every brand has learned a lesson the hard way. Everyone.
There's enough bad content out there or inappropriate content that they don't want to be associated with. It's a really hard challenge, especially for UGCs. It's not because they're being negligent, it's because their challenge is hard. UGC is very difficult to police and monitor. If you look at the creative acceptance policy and essentially the content police at Facebook and look at every single policy that they have to carefully scrutinize to figure out what's permissible and what's not, it is a very difficult challenge that for the most part, premium content doesn't have. We feel like we're in this really great position that especially in TV, where there's the most amount of sensitivity around running your premium ad next to content that isn't great, that we're partnering with the very most premium content.
Again, that doesn't mean This Is Us, but there's not much risk in associating with Major League Baseball, while it probably is billions of user-generated 15-second videos.
Hi, it's Mark Zgutowicz with Rosenblatt. Just wanted to follow up on a question regarding take rate, then also had a question just on your success with CTV vis-à-vis Amazon and Google. First on the take rate. I was hoping you could maybe balance sort of your guidance, the 15%-20%, which you've consistently obviously exceeded the top end of that over the last couple of years, with sort of what the dynamics are within the take rate itself today. You obviously have media and you have data. The underlying question is, you continue to sort of beat that sort of number despite media sort of rates coming down. The presumption is you'll continue to remain above that sort of 20%.
If you could maybe talk about the components and sort of how that might be different than we've seen over the past couple of years as we look forward.
Sure. The thing that we care most about is grabbing land. The way you do that is, of course, operate a profitable operation so that you can reinvest that money as much as possible. Then you also have to look at where does the pressure come on your take rates? One place that it obviously comes is from the value that you add. That's the one that I think about most. We talk about the economic concept of consumer surplus, which is when you're giving more value than you're charging, that surplus is consumer surplus, or just what's economically defined as consumer surplus. That is the thing that makes your customers love you. That is the thing that has made companies like Netflix just loved.
There's a cult following behind them because you essentially say, "I can't believe I get all this for $10." That's how I feel with Spotify, like, "Wow, I can't believe I get all music for this." Their model is built on, "Well, I need everybody to subscribe or else we don't make any money." It's a really beautiful thing that they're shooting for creating a consumer surplus. I look at it as we charge this certain rate, and we're adding the most amount of value in the value chain today. We don't have to change anything unless competitors make us. Competitors have not really done that to date. It is true that sometimes the big guys will offer discounts or rates that are lower than ours, but I don't believe that ever makes up for their lack of objectivity.
What they tend to do is make it up in the media rates that they funnel to their own owned and operated inventory. Because of that, we don't really see any pressure or legitimate pressure that can come from anybody other than independents. I think we've been a little bit on the high side of that, mostly because of the lack of competition. I don't expect that to last forever, which is why I do think that we'll come into that range. I do think that that's defensible, and as I've always said about our business model, if that ever has to change, I don't expect it to change so much that we have to change our business model. As an investor, that's what I would be worried about.
If there are meaningful changes, do you have to change everything about your business model? No. As long as we stay in that range, we keep doing what we're doing. If we're on the low side of that range, it makes it that much harder for there to be very many competitors, and we will have grabbed more land. If I can go faster by lowering the rate and ultimately make more money as a result, I'll do it. We just keep looking for those opportunities. To date, we think we're as close to equilibrium as we can be given the state of the market.
That's helpful. On the CTV question, you've had a pretty clear runway, if you will, in terms of being out in front of Google and Amazon, clearly in the marketplace in terms of share. I'm curious, as they sort of get their act together or maybe their technology starts to move up more comparable to yours or at least acceptable for a content provider to monetize their inventory, how do you look at sort of your share of CTV and in terms of how that might affect your trajectory that you've seen so far, which has been sort of off the charts if we look over the next year or two?
I'll take the first side of it, Brian, if you want to add anything, or Tim. While I believe our technological advantage in connected TV is a huge reason why we've been winning, I actually think the bigger reason is actually the objectivity and the thing that you have to keep in mind in the psychology of TV. Content owners are reluctant to partner with Google, Amazon, and Facebook. Like Disney asked Jack Dorsey and Sheryl Sandberg to leave their board, what, two years ago. If you're Jack, you're saying, "I don't even make content. How am I your competitor?" That's how concerned they are about owning and controlling distribution in the future. There is a serious reluctance for content owners to be overly dependent on those companies.
It is why Netflix had to pivot their business model to be a content-making company instead of the distribution company that it was four years ago. To me, that is the big story here, is that there is a bias for them to work with us because we're independent. That ranges from inside the U.S., outside the U.S., and all three or four categories of the content owners or content monetizers that don't want to be too dependent on Google. To give one insight here, there are content owners in Germany. Part of the reason why our numbers, those of you that follow closely, we've talked about how strong our numbers were in Germany almost every quarter in 2018.
That is in part because on the connected TV front, the equivalents of NBC said, "We're not going to give any access for demand to Google." While Google on YouTube has said, "We're not going to give anybody access except for DBM," they've done the opposite, which is, "We're going to give everybody access except for Google." That is because they don't want to create any more dependence on them than they have to, and their commitment to owning their own content and distribution. That's much higher than our technological advantage in what has put us ahead, and we're just using all the profit and all the head start that we have to then go accelerate the technological advantage.
Yeah. The only thing I would add is Tim talked about going fast and being first with new supply partners, and connected TV is one space where we've gone fast, and I think we are kind of first to market. We don't want to be first or the only one forever, because ultimately, if a network or MVPD or whoever is selling inventory to us, if we're the only source of demand, that's not a long-term sustainable business model for them. I do think there's reluctance to partner with Google and Facebook, but whether it's our other competitors like MediaMath or DataXu or whoever, I actually want them to come to the table and bring more demand because that'll make the networks and MVPDs put more inventory. It'll become more of a long-term virtuous cycle. We want a head start, and we want to try to maintain that.
At a certain point, we're comfortable just winning on our merits every day against platforms like we always do.
Well said.
To stay on time, we're going to have one more question, then we'll move out to the reception area. Asim.
Thanks, Chris. First, just want to say thank you and congratulations to you guys. Fantastic 2018. Two quick questions. One was for Dave on these Megagons. Now that you have the capacity to do eight of these, what are those eight? What are the four? Whatever the number is.
Yeah.
The second question relates to offline measurement. What are you guys doing in offline measurement and attribution? That's something I didn't see. Saw a checkbox in the demo, and how you're partnering for that. As it relates to maybe a third question, if I may, on self-serve. If you look at the bigger platforms, one of the things they did really well was how they made self-serve available to SMBs. Is that a part of the roadmap at some point, or is this really not going to be for SMBs in terms of product evolution over a longer period of time?
Asim, this is one we have to divide and conquer. You got question one, two, I'll take three.
Okay. Yeah, Is that working now? Yeah.
No. There you go.
It was working. Okay, just can't hear it. The Megagon was very much one release along a continuum on user experience. One of the primary places that we're pouring in resources is finishing that thought and layering in all the additional AI, getting to the place where we're happy with the division of labor between people and machines. That's really big. There's a lot of other things we're doing. We're trying to get forever better at multitasking, and every year we talk a lot about how many things can we do this year, and that's gone from one to three to seven or eight things that we can go after. I talked about some here. Some aren't ready to talk about. We're also investing a lot in measurement, is something that we're thinking a lot about.
There's a lot of different angles that we're pursuing, but all just basically trying to get better at going from 10 million requests a second to the right 20 every second for your brand. We can get better at that for a very long time. I'm really focused on just continuing to widen the gap between us and the competition on that single issue.
On offline measurement, that's been an active product for a couple of years. We've got about 10 or so different partners built in the platform. If you want to measure CPG sales with Oracle or Datalogix, at the click of a button. If you want to measure store place visits from Factual or companies like PlaceIQ, you can do that. If you want to measure car dealership sales, we have companies for that. We're working on partnerships in movie ticket sales and other sectors. That's an active part of the platform and absolutely a place we see growing quickly and see more advertisers moving to.
On the self-serve, it's something we spend a lot of time thinking about, talking about. We've debated it for literally years. The debates go something like this. We acknowledge that most of the success of Google and Facebook have come as the byproduct of them building easy on-ramps. Facebook has, I think, done the best job in the history of advertising, of making it so that you can sort of part ways with your advertising spend in 90 seconds. The on-ramps are so easy. I don't mean to say that like it's not effective. It is. Those on-ramps are the most impressive thing that I think Facebook has done on the revenue standpoint or from a revenue standpoint. All of that is built on the back of their consumer surplus. It didn't cost me anything to connect with all the people closest in my life.
The fact that you're building on that human connection is something that is just really amazing that Facebook has done, that Google I have all information. We're talking about the consumer surplus of Spotify. I have all music at my disposal for $9. With Google, I have all information at my fingertips for nothing. Seemingly nothing. It is just so impressive. That consumer surplus makes it as SMBs, people will just trust them in the millions. Millions of people give them budgets. That, to me, is the most impressive thing about their business models. Now, that said, there's the rest of the internet that we can objectively help them buy, and we think all the time about the fact that what if we could tell consumers about the objectivity?
What if we could help them do on the rest of the internet what they've been doing on these specific destinations? The difficult thing is we don't have the same relationship with consumers. You have to work really hard to make certain that the phone doesn't ring with every tiny little budget that is asking, what happened to this? Can you explain all the complicated things that we've been talking about all day to a plumber who just wants to know how his $150 was spent this month? All of those challenges is what has made us continually say, we're very good at one thing, which is helping the biggest brands in the world objectively and intelligently spend. We're not anywhere close to being done with that.
Given that of that $1 trillion, we think that roughly half of it is going to be spent by people that we consider to be big. There's a massive TAM in all of that. We're not done thinking about is it possible for us to do that. We're always thinking about what is the Next Wave.
The only thing worth adding to that is we have some customers today who have built our APIs who do tackle that market. There's some large newspaper companies who have huge sales teams. They get the $100 a month campaigns or from the plumber, the car dealer, whoever, and they do run those on our platform. When you see that number of the close to 50,000 advertisers, there's a bunch of SMB in there. We don't do it directly, but whether it's a company called Choozle that has built a UI for small business on top of our bidders. They've been very public about that. We see some of our customers tackling that space, but we just haven't tackled it directly.
In fact, one of our former board members, when he came off the board, immediately invested and became part of a board of one of the companies doing that very thing, because he sees that as a huge opportunity. We're already capturing some of that, and that may be enough. We'll keep watching. Awesome. I just want to add just a little bit of punctuation to just the entire day. First of all, I know it's a long day for everybody to come out, and this group of people put a lot of time in to putting together presentations to share and give insight into our business. I hope that a little bit of our passion shines through, or at least a little bit. I promise you it wasn't all of it. It's a different setting than where we're usually presenting.
For us, this is something we think about night and day. It's something that we believe we're looking at a once in a lifetime opportunity. That there is this transition that is happening in media, and we really do. This is the way we honestly inspire our team as well, is we remind them that the world of media is better because we're in it. Because if we continue to maintain our objectivity and continue to make it possible for advertisers to spend on the rest of the Internet, we make it possible for the rest of the Internet to continue to innovate, to create great content, and we make it possible for the brands that would otherwise be replaced by others to continue to be competitive. We think we make the marketplace way more competitive, that the world is better with us in it.
We just can't thank you enough for your support, both in just supporting the vision, but of course, as investors, it's the only thing that makes that possible. So many, if not all of you, are a part of that, as the way we see it. We just want to say thank you for that.