In the interest of time and for the webcast, we are going to get going. It is my privilege to kick off our first fireside chat of day two with Bill Ready, the CEO of Pinterest. I am going to read a quick safe harbor, and then we are going to get into this. Some of the statements that Pinterest will make today may be considered forward-looking. These statements involve a number of risks and uncertainties that could cause actual results to differ materially. Any forward-looking statements that Pinterest makes are based on assumptions as of today, and Pinterest undertakes no obligation to update them. Please refer to Pinterest's latest Form 10-Q and Form 10-K for a discussion of the risk factors that may affect its results. Okay. Bill, thanks so much for being part of the conference again this year.
Thanks for having me.
It is always great to have a chance to talk to you. It has been a few years now since you joined Pinterest as the CEO. I think the anchoring theme I always consistently hear from you is about making Pinterest as a platform more actionable.
Can you talk about the journey you have been on measured against that goal?
Yeah
from when you came into the company?
Yeah, absolutely. The headlines, Pinterest has become an AI-driven shopping assistant for 640+ million users, becoming one of the fastest-growing platforms in social media, consistent double-digit growth in users, strong engagement growth. At the core of that has been the actionability that you just mentioned. Coming into Pinterest four years ago, it was pretty much a platform for visual discovery and browsing, but it wasn't a place to buy things. It's now the case that more than half the users that come to our platform are there to shop. Gen Z is the largest, fastest-growing cohort, more than half our users, and primary use case is shopping.
At the core of that is making it so that you can discover more of what you're looking for, make more of those purchases, and that has been at the core of the revitalization of the platform, both in terms of the user side of it, where we're getting much deeper engagement, particularly with those Gen Z users. Stated simply, Pinterest is where Gen Z goes to shop, and it's a place that they're thinking of as a first place to go search increasingly as well. There's an Adobe study that we've cited previously that 39% of Gen Z thinks of Pinterest as the first place to go search. Why? 71% of Gen Z sees Pinterest as more personalized than other places to go search.
At the core of it is it's not just us using AI, it's us using really unique data that we get around the human curation on our platform. AI doesn't have style and taste on its own, humans do, but that's what people do on our platform is curate that style and taste, and we've used AI to really superpower our recommendations to users, giving them great recommendations before they even know what they're looking for, that users will say, Oh my gosh, Pinterest just gets me.
Yeah.
And increasingly make it so that they can shop and buy. That is why now you see that from a platform that was almost entirely upper funnel ads four years ago, we have talked consistently about two-thirds plus of the business is lower funnel performance. We talked about more than 5x the number of clicks to advertisers over the last three years. Those things all get to the deep actionability on the platform. We have had great progress there. At the same time, in the course of monetizing a platform like this, we are still relatively early on. There is still a lot more opportunity. One of the strongest parts of the story has been our growth and user engagement, and then the monetization, I talked about the shift to lower funnel. We have made great progress on that, especially in our home market here in the U.S.
But then international, where we have more than 80% of our users, still roughly 20% of our monetization, so a lot more of that to go there. When I sort of look at the progress, we have clearly proven out Pinterest as a shopping destination, clearly proven out we can make it actionable, clearly proven out that we have differentiated signal to train the AI on to give fantastic recommendations to users so they want to come back and shop more. That is a global phenomenon. Then the monetization always follows in behind the user behavior, and so we have made great progress in the U.S. We are translating more of that international, but there is a lot more of that to do.
Okay. Building on that answer about your future vision for the company, we got the news in the last week of Julia's departure from the CFO role. Can you talk to us, when measured against where you want the platform to go, what you will be looking for in a new CFO?
Yeah, certainly. Having built from zero to many billions multiple times, these things are like relay races, right? You think about what's the team and the things you need to do for each leg of the race. I touched on this a little bit in the last comment of what were the things that we needed to address? Four years ago, this was basically a single product platform in a single geography, and we're now moving into multiple products, many more geographies, much greater complexity, and we've been building up a team for that leg of the race.
If you look at what we've been doing more recently across the team, it started with a new chief product and technology officer that led shopping at Google, has been doing a fantastic job with what we're doing with the tech platform, bringing a great team there. Recently, a new head of product engineering for ads, a new head of infrastructure, the new head of the internationalization of the platform on the product engineering side, new chief business officer, new chief marketing officer, and so we're really building up the team for where we need to go next. Similarly for the CFO, when I think about that, we have now proven out that we can monetize our audience beyond just one surface. We did the acquisition of tvScientific.
We've got multiple products within Pinterest now monetizing beyond the one surface and going to many more geographies. So when we look at that complexity of the next leg of the race, we've been building a team around that, and these are things that we'll be solving for with our next CFO, in that we'll want a deep operational partner. Grateful to Julia. She ran her leg of the race well. Very good partner through that, and I'm sure she'll be great in the next thing that she goes and does. But if you look at where we were four years ago, there's a lot of what we were doing that was addressing a lot of things that ideally probably should've been done before the company was public.
We were really getting a lot of that operational rigor and hygiene, and now we're in this place of expanding to more products, more geos. That brings greater complexity. So, we'll look for another great operational partner, but one that has done a lot of that kind of operational complexity at scale and across the rest of the team. That's what we've been building for, and we're seeing really great talent, really great candidates for our CFO search. So I'm quite confident that we'll get a really exceptional person in that role that will be a great partner for the next leg of the race.
Okay. There are so many themes of convergence that we have been talking about over the last 12-18 months, search, social media, commerce, AI. When you think about the broader discovery process that consumers are going on right now against those themes of convergence, how do you think about positioning Pinterest against consumer discovery over the medium to long term?
Yeah. Well, I touched on this a little bit that we have made tremendous progress. I think it has been the strongest part of our story is the. It is a big change in the story. Four years ago, Pinterest was a platform that was rapidly declining and users had missed the next generation. Now to a platform of 12 straight quarters of record high users. Gen Z is more than half the platform, our largest, fastest growing audience.
We have talked about many times how we are growing across every geography that we track, across the generations that we track, and so it is broad based. At the core of that is people are thinking about different ways to engage in their shopping experiences than what they did previously. I have said this before, that the first 25 years of e-commerce in a lot of ways sort of solved buying, but killed shopping.
The distinction being that if you knew what you wanted, then you wanted to get it the cheapest and the fastest. There are a lot of things built for helping you do that. But a lot of the rest of the shopping journey was about, well, I do not know what I want to buy yet. It is the sort of I will know it when I see it problem. If you think about how people shop in the real world, that is a lot of it that whether they want to update the wardrobe for fall, or they are thinking about what to make for dinner that night, they will go into a shopping setting with a sort of loose idea. But, if you ask them, Hey, tell me exactly what it is you want to buy, they say, Well, I do not know.
I will know it when I see it. That is what Pinterest is solving, but we are solving it now in a way that is going from that upper funnel discovery all the way, basically sort of dream, decide, and do. So what is that initial kernel of the idea? How do I refine that? How do I make a decision? Then how do I easily take action on that? We are solving across all of those.
And a thing I will share, one of the questions that we will often get, especially on the strength of our Gen Z usage, and it is funny, I used to get this question all the time in the early days of Venmo where people say, Are these millennials ever going to spend money? And I was like, Of course, they are going to graduate college and they are going to spend money and all these things. Then we get that question about Gen Z a lot of like, Well, can you monetize those users? With Gen Z, not only are they our largest, fastest growing demographic at more than half the platform, they have the deepest engagement, the highest retention. They are also the most AI native. As we are building AI native capabilities and things like that, they are deeply versed in those things.
Our assistant capabilities we put out there, deeply versed in those things. But Gen Z on our platform, we now see the ARPU of our Gen Z users in the U.S. is roughly consistent with the ARPU of our older generations like millennials. I think that bodes really well for the future monetization of those Gen Z users that, of course, they are going to shop. They are shopping. They are shopping quite a bit, which is why their ARPU is now consistent with the next generation up with millennials. I think that bodes really well for the future monetization of our platform because the purchasing power of those Gen Z users is only going to increase, and they see Pinterest as a place for them. Again, Pinterest is where Gen Z goes to shop.
We see that even as their purchasing power is early in its growth, they have already equaled the ARPU of millennials on our platform. We had not shared that previously, but that is what we are seeing on the platform, and it is exactly what I would have expected having built commerce experiences previously as these generations mature. But the depth of the engagement with our Gen Z users is just quite significant and outpaces the other generations on our platform.
Okay. Understood. You spent some time on the Q2 earnings call talking about your approach to AI as a company. Against the landscape of what is available to you, models, products, things you can bring into your ecosystem, talk to us a little bit about your AI strategy, both in terms of what you can pull into your ecosystem from the AI landscape, and then how you can take elements of your data and build personalized, unique experiences at Pinterest in a cost-efficient manner.
Yeah. This is from the time I came over from Google. I have spoken about this consistently that in this coming wave of AI, the sort of model capabilities will be available to many. There was going to be a lot of opportunity for places that had unique data that AI thrives on feedback loops. So what are the places that had unique data, unique user feedback loops? For Pinterest, I talked about how we are using that unique human curation on our platform to train our AI to give really distinct recommendations. That is at the core of the re-acceleration of the platform that basically everything you see on Pinterest is served by AI and has been for a couple of years. That is one of the first things I thought about doing coming into Pinterest.
But we use a combination of compact fit-for-purpose models and taking open source, open weight models and post-training on our data to really great effect. We have been talking about this for years. Now there is a lot of discussion of open source, open weight. We are very early on that. There is a lot of discussion of model routing. We are very early on that. There is a lot of talk about, oh, well, maybe you do not need the frontier for every task. Maybe that is over-solving the problem at much too great an expense. Well, check the tape.
We have been on this for several years, and it is part of why, even as we have made it so that everything you see on Pinterest is served by AI, we have also had significant expansion of margins in our business over that period of time because we have been able to not only align those AI use cases with highly monetizable events that are very commercial. It is also the case that we have been able to do it at very good cost effectiveness, and that is because we use compact models trained on our unique data. We get great results from post-training open source, open weight.
I think that is a really important thing as you look forward, that in the initial sort of wave of adoption, everybody just racing to make sure they did not miss out, and particularly less sophisticated companies and things like that were just like, Hey, just give me whatever I can adopt. As you now see token bills coming in and things like that, people are thinking a lot more about, oh, I probably should think about the right tool for the job. Using the frontier for everything might be a little bit like using a Formula One engine to mow your lawn. You may have not only massively overpaid, you may not have solved the problem as well as something that was built fit for purpose for the thing that you are trying to do.
I think one of the most exciting things that has happened in the industry is the advancement of open source and open weight. I said on our last earnings call that, I will say it again because I think it is really important, any CEO that is not heavily leveraging open source, open weight models is almost certainly wasting a lot of shareholder money.
Yeah.
Because you can get similar capabilities at dramatically lower cost. For us, we shared that when we use open source, open weight compared to a comparable size model, we are able to get cost at less than 8% of what the comparable closed models would be. We also see greater effectiveness because you cannot train the closed models on your own data. So when you can train on your own data, that is greater effectiveness, and it is actually greater safety as well because you can run it in your own secure environment, which you cannot do with the closed models. So you know where your data is going with the open weight models when you are running in your own environment.
Not only do you have a very good thriving open source, open weight ecosystem now, including here in the U.S. with U.S.-based models you have the hyperscalers leaning in and doing what they have always done. If you look at how important open source software was over the last 30 years, a big part of that was that hyperscalers came and sort of packaged up open source to make it easy for others to use. Most of the internet runs on Linux. Well, Linux, most people would have thought of as sort of harder to use earlier on. But then you had hyperscalers sort of package that up, make it really easy for others to consume.
The same thing is happening with models now that you have these open source, open weight models that are just only a few months behind on any comparable size, tend to be only a few months behind the frontier. Then you can post-train on your data, get better results, much lower cost, and now much easier and safer to do because you can do it in your own environment. Whatever hyperscaler you use, you can do it in the environment where your data already exists, so you know your data is secure. It is not going out someplace else. So I think that is a really important trend, and it is a good thing for the thriving of the overall ecosystem just as open source was. Most of the companies in the Valley wouldn't exist if not for open source software.
I think that just says that this sort of world-changing technology will be in the hands of the many rather than the few. The innovation will be for the many rather than the few because of what is happening with open source and open weight and having chip makers like NVIDIA really behind that, having hyperscalers behind that I think just says there is clearly a good ecosystem there. It is what we have been doing for several years now of saying right tool for the right task gets you better effectiveness, better cost, and you can do it safely and securely. I think that is a really good thing for the broader tech ecosystem.
Okay. I wanted to turn to the advertising landscape. We are sitting here in early September. What are you seeing in the ad market today? Away from sort of the environment, how are you also thinking about the competitive landscape and where Pinterest fits into the broader ad ecosystem right now?
Yeah. As I get started on this one, I want to make sure I am really clear that I am not updating or addressing guidance with this. It is not our practice to give intra-quarter updates on trends. That said, the puts and takes are consistent with what I talked about on the earnings call. I will spend just a minute sort of going back through what we talked about on the earnings call with regard to the puts and takes in our business. We have talked about the user growth a bunch. One of the really exciting things in H1 of this year is the re-acceleration of our UCAN business. There was a lot that we were working through with adjusting to tariffs last year, being a shopping destination, and sort of the impact on retailers. That was a lot that we were working through last year.
The re-acceleration of our UCAN business is quite noteworthy through H1. We talked about how the AI-driven ad platform, our ROAS improvements for advertisers, that we felt like those things were durable as we looked into Q3. That is something we are really excited about. We also talked about, though, on our international business that we were seeing pressure there and on two fronts. One, new regulation in Europe that puts limits on Asia cross-border sellers. So technically not tariffs, but you can think about it similar to what happened in the U.S. with tariffs, where EU is putting up new restrictions on Asia cross-border sellers. That is a new thing in Q3. That creates pressure in the international business. The other thing is just we are restructuring on the international business with our go-to-market, just as we did in the U.S.
You saw in the U.S. that as we re-accelerated, it was the AI-based ad platform improvements as well as significant overhaul of our go-to-market and some of the early signs of benefits of that. We are basically taking that playbook from the U.S. of shopping in the U.S., the go-to-market in the U.S., and now deploying that internationally. We talk about, okay, that will create some near-term pain to restructure those things, but that we think is very good for us in the medium to long term, just as it was in UCAN. Those were some of the puts and takes in terms of what we see in the business consistent with what we talked about on the call.
Stepping back from sort of these near-term things like new regulation in the EU and sort of short-term restructuring of our go-to-market and things like that, the shopping behavior of our users is a global phenomenon. The AI-driven platform improvements are a global phenomenon for us. We continue to be very excited about that because what we are proving out is that not only can we drive great actionability, great outcomes for advertisers, we are creating a truly full funnel experience.
Where historically you would have upper funnel, mid funnel, lower funnel happening in different consumer surfaces, we are bringing that together and we are proving out that not only can we drive great lower funnel performance, but as advertisers are seeing more and more that, well, sort of discovery and research may happen in one place, the last click may happen in another, our ability to tie that together and drive great performance is really shining through with advertisers. I had shared previously that it is what great CMOs always knew, that the last click was not the only thing that mattered, but that when we had advertisers doing multiple objectives, thinking sort of upper and lower funnel together, that was two times the conversion on that. Well, that is now three times the conversion for us.
Yeah.
That benefit of doing multiple objectives with us because we can span upper, mid, and lower, that is now three times the conversion. We see that as we get the sort of flywheel of more users curating their taste on the platform, our AI-driven recommendations get better and better off of that signal, more and more differentiated off that signal, drives better and better outcomes for advertisers, and lets advertisers engage across discovery, sort of decision making, and then ultimately purchasing. We think that is something that is quite unique. As we talked about the re-acceleration of our UCAN business in the AI-driven ad platform improvements, retail is a strength for us, right? Shopping is a strength for us. We think that is something that bodes really well.
As we are rolling out new AI-driven capabilities to consumers, I think we are striking a great balance between the AI sort of powering their experiences, but in a way that the users feel like is very human-centric to them. As there is discussion of AI slop and all these things, Pinterest is a place where users actually get to directly engage in how they are refining their style and taste, and the AI is there for the assist.
Yeah
It is not sort of overpowering the experience. I think that is not only good for users, but advertisers are really liking how we are driving great results, but in a way that is brand authentic.
Yeah
to them across all those stages of the funnel.
Maybe just one quick follow-up on that. When you think about where you are going over the medium to long term, has your view with respect to agentic commerce evolved at all? Because I think that has been a big talking point over the first day-plus of the conference, because you are trying to get to more shoppable, you are talking about conversions.
Yeah.
But what role does agentic commerce either on platform or off platform have to play in that?
Yeah. We have had some really exciting progress on this. I have talked about how Pinterest has become an AI-driven shopping assistant. We took a different approach than others. In the same way that I have said the first 25 years of e-commerce sort of solved buying but killed shopping, as you saw a lot of others go after agentic commerce, they went after the buy first. I said very consistently that clicking the buy button is one of the easiest things, not only for the consumer, but also technically one of the easiest things. The real promise of agentic and where consumers wanted the help the most was making the recommendation before they knew what to ask.
That promise of agentic, of the agent sort of out there shopping for me before I even know what to ask, before I know what to do, or I was looking for something and I come back the next day and like, Oh my gosh, I found the things that I was looking for. This is what we have been doing on our platform, what we have been focusing on, and what is driving the depth of engagement. We have been consistent on that. It is at the core of how we are driving engagement, and we continue to progress on that.
Our new assistant capabilities, where basically you can think about we have had AI in the background on a lot of these things, like making all the recommendations, every pin you see is served by AI, the really powerful personalization experiences all being driven by AI, and that was in the background.
We have now brought that more to the foreground. Our new assistant has been made available to virtually all users in the U.S. And we are seeing some really exciting things there in that in the same way that we took Pinterest deeper into the lower funnel with actionability, with clicking and purchasing, it is also the case that we see that we are driving so much discovery and decisioning that those Gen Z users on our platform, they have all used AI, right? They have all used chatbots. We would see that they would start with Pinterest and then, oh, they would find it on Pinterest, they would purchase on Pinterest, but they maybe have a research question of, Oh, hey, I want to buy that pair of shoes, but I have not bought that brand before.
Do they run big or do they run small? Well, now with our assistant, you can just ask that right inside of Pinterest. The encouraging early signs we're seeing there, we've talked about how on our platform overall, we now see over 80 billion searches per month, and more than half of those are commercial, which is a much greater commercial skew than you would see in other places to go search. We are clearly a shopping-centric destination. With our assistant, we're seeing that while searches overall for us, more than half would be commercial, with the assistant, it's 80% commerciality.
Back to us aligning AI with highly monetizable use cases, that continues to play out as we bring those assistive capabilities more in the foreground in a way that still is very visual first, but lets you ask some of those follow-on questions right inside the platform in a way that's really effective and very cost-efficient. It's notable, we said this on the call, that at the same time, we made our assistant capabilities available to all of our users in the U.S., or virtually all of our users in the U.S. We also took up margin guidance, which gets back to using AI in a way that is very effective because of our unique signal, but also in a way that is very efficient from a cost perspective.
Okay. Super helpful. We have a few minutes left. I just want to try to get through maybe one or two big picture topics. One we've written about and we've talked about on our earnings calls is the evolution and the scale you've been building around Performance+.
Can you talk a little bit about Performance+, how it fits into your broader advertising ecosystem, and how to think about what that does for the platform over the medium to long term?
Yeah. This is a place that we have been very excited about, and we have seen really good results with advertisers. We were setting out on that journey after some of the larger platforms, but we have seen really, really good progress. We only went GA with Pinterest Performance+, our AI-driven ad platform, at the start of last year, a year and a half into GA, and still adding lots and lots of features. I think for us and other platforms, we will be adding features to this in perpetuity. You should not think about it as, oh, there is a launch, and then did you get everything out of it from just that one launch? It is like you are building on that in perpetuity. But we shared previously 30% of our lower funnel revenue now going through Pinterest Performance+.
The pace of that adoption has been quite exciting, and I think when you compare it to the pace of adoption on the larger platforms that were at it for multiple years before Pinterest, I think that pace of adoption compares really favorably. I would say almost roughly half the time of what that would have taken from the public commentary of other platforms. So we feel really good about the pace of that adoption and the fact that it is helping advertisers tap into really unique shopping behavior with really good incrementality. Again, meeting those users in a place that is not only letting the advertiser get great results, but also really tell their story in a way that differentiates.
I think this is something that every advertiser is struggling with is as traditional SEO-based web traffic declines, which you can see in all kinds of data out there that that is declining and more of these decisions happen other places. As those decisions are happening in other places, how much can the advertiser engage in, advertisers, they have been doing a lot of things to try to figure out like, well, if I do this thing over here, does it eventually some way trickle into something that happens on the other side? In a lot of ways, you are competing to be the salt in the soup on that. How do you? For us, we are making it so they can engage directly, and they are not competing to be the salt in the soup. They get to be the main course.
Yeah.
They get to have their brand show up exactly how they would want it to show up in a way that is actually really compelling for the user also because since we are a shopping destination, ads can be great content on our platform. These things, I think again, we are quite excited about that in the rollout of Performance+ is just making it easier and easier for advertisers to take advantage of. We talk about cutting the time in half for them to create campaigns while also getting these really great results in terms of how to meet the user, where to meet the user, how to engage with them across multiple stages of the journey, and then having the AI just figure that out for them versus them having to go do all that manually.
Okay. You also decided to acquire tvScientific. Talk to us a little bit about how you are thinking about that asset now that it is inside your portfolio, amplifying some of the commercial proposition of the advertising offering, how it can improve signal, how it can drive more performance outcomes.
Yeah. The basic premise with this, tvScientific is a Performance+ advertising platform for connected TV. One of the things that we have talked about is that as we made Pinterest a shopping destination, that job one is certainly make Pinterest itself a destination, which it clearly is. More than 85% of our usage comes to our mobile app directly. Pinterest is a destination. There is a lot more to go on that. The audience that we have, I would say, we believe we have more direct knowledge of user commercial intent probably than any other platforms aside from Google or Amazon. What you have seen in those platforms is that you can take the value of that audience to build great ad products beyond one surface.
We have taken our highly commercial audience and our knowledge of tastes and preference and what users are looking for, and we are using that to show much more personalized ads on connected TV, and we are proving that in the same way that on our platform we said doing multiple objectives leads to three times the conversion rate when you do sort of upper and lower together, we are now showing that on connected TV. The historical world was like you and I would watch the same football game on Sunday, and we see the same ads. We may have some similar interests. I am sure we do. We get along pretty well. I am sure that at any given moment, you and I are shopping for different things.
Yeah.
We should see different ads, and that is not just better for the advertiser in return, but the advertiser is better for the user. With tvScientific, it is letting us show those much more personalized ads. The thing that we have talked about is that as we do that, we are able to show that adding our audience on top of what tvScientific was doing with Performance+ advertising and CTV, there is a more than 60% lift in purchases.
Yeah
from doing that. How valuable is that audience? Tremendously. There's a lot more that we'll do in connected TV, but we think the value of that audience is quite noteworthy and lets us think about how we monetize across multiple other surfaces. We're at the very beginning of that. We think that's a big forward opportunity, and maybe just a place I'd say in terms of where we are in this moment, AI is moving very quickly and where people think the value's going to accrue keeps shifting. There was a stage of this where people were saying, Oh, all the value goes to the model creators, right? There will only be two or three companies in the future because all the value goes there.
Then, Oh, well, wait a minute, hyperscale's going to be pretty important in this, and Chips and memory are going to be pretty important in this. Now you see with open source and open weight, you have a lot of commoditization of model capabilities. In the same way that open source software didn't lead to zero proprietary software, in fact, most proprietary software, much proprietary software leverages open source, you're going to have a mix of both closed models and open models. That says that with the advent of open source and open weight models, that's going to allow for much broader value creation. A lot of people have been focused rightly on there's a lot of infrastructure build-out and those kinds of things.
I think the next part of this that people haven't paid enough attention to is how much value there is in distribution, how much value there is in unique data. This is a place where at 640 million users, we are clearly one of the largest practitioners of applied AI, bringing AI-driven experiences to a huge mass of users in a very unique way. Search is in this expansionary moment, right? It's clearly the case that chatbots are a new form of search for most consumers. That's expanding the search market overall. Search has also been fragmenting for years in that there were a few general-purpose ones, but many vertical-specific ones. We're driving that sort of vertical specific around shopping and visual, but we think there's a lot more of that that we can do across multiple surfaces.
We're in a moment where I think that people are seeing that, okay, yes, there'll be a lot of value in the closed models, but open models make it so that a lot more value can accrue to those who have distribution, who have consumer relationships, and those who have unique data, and tvScientific is a first step in us proving out the value of that audience across multiple surfaces.
Okay. I think in the interest of time, we're going to have to leave it there.
All right.
I always love getting the opportunity to chat, Bill. Please join me in thanking Pinterest for being part of-