Michael just said he feels special that you guys showed up, even though you're probably fried from the day. So, and then just one housekeeping item. We are going to stay a little longer here because this is the last session of the day. So at 5:10, when this session is officially over and the webcast is done, whoever wants to hang around for, like, 20 minutes for some additional questions, feel free to do so. We're not going to go up to the breakout room because it's not necessary. With that said, I'm Jason Ader with William Blair. Very pleased to have Michael Gordon, COO and CFO of MongoDB, and Serge Tanjga, SVP of Finance. Regulars at this conference now.
Yes.
Appreciate it.
Great to be here. Thanks for having us.
Before we begin, I'm required to inform you that a complete list of research disclosures and potential conflicts of interest is available on our website at williamblair.com. We're just going to do a fireside chat format here, and hopefully, we'll have some time at the end for Q&A. But like I said, at 5:10, we'll cut the webcast, and then we'll have a little bit more of a kind of interactive discussion. To start out, Michael or Serge, I think maybe you're, because Michael's voice is going, you might want to do this.
I apologize. I'm losing my voice.
For investors less familiar with Mongo, I'm sure most people here are familiar, but we may get a few that are not super familiar. Can you just give a quick history lesson on the company, and tell us what pain points you solve for customers?
Excellent. I'm happy to start.
Yeah, sure.
So we play in the database software space. That's a $80+ billion market, as per IDC last year. One of the largest markets in software, and then also, interestingly enough, still a very fast-growing market, still a double-digit growth market. And the reason why that dynamic is in place is because of many of these things that you've heard over the years, such as every company is becoming a software company, software is eating the world, competitive advantage in every industry is increasingly dependent on the software experience that they provide to our customers. And that market is dominated by a 50-year-old technology known as relational or SQL. It's literally 50 years old. I think the original paper that created it was either 1969 or 1971.
So a technology built before mobile, built before internet, obviously before cloud, and obviously before AI. And we, among others, were founded roughly 15 years ago to address some of the challenges of that technology that became increasingly clear to developers. And those are. The legacy technology is not terribly developer-friendly. It's hard to work and reason about and doesn't work compatibly well with modern programming languages. And then generally, the second bucket is performance. It's hard to scale. It's impossible to scale horizontally, so it's expensive to scale vertically. It's hard to create performant and globally distributed applications, and as applications that we're used to become more and more demanding, these challenges become more and more apparent to developers and ultimately to the customers.
We were founded in 2007 because our founders were developers who were working with relational databases and realized all these challenges and realized that that's not a tenable state of a large industry in the long term. We were founded on a completely different paradigm, known as the document model, which solved the pain points of the customers that we just described. We've been around for 15 years, went public in 2017, did roughly $1.7 billion of revenue, and guiding to roughly $1.9 billion this year.
So call it roughly 2% share in this market, and that's sort of fundamentally what gets us excited about this opportunity, which is that despite the amount of success that we've had so far, the traction that we have in the market, the opportunity ahead of us is still enormous. We think of ourselves as the general purpose solution, i.e., all but some of the nichiest use cases can and should be built on MongoDB, and we have this great opportunity to continue growing and gaining market share over the years.
Okay, great. Thank you. And then I'm sure on everybody's mind, what happened in Q1?
Yes. So thanks for the question, Jason. No, so for those who know, we reported results last Thursday. It was a more mixed quarter for us. That's not normally our pattern, and so there are plenty to talk about, particularly to call out sort of some of the unusual things. But maybe just to run through some of the summary key metrics, 22% year-over-year growth on a top line. Our database as a service offering, which we call Atlas, grew at 32% on a year-over-year basis. That's now 70% of the total business, up from, you know, virtually nothing at the time of the IPO, just to give a sense for folks.
We beat the high end of our guidance, although, as I'm sure, we can get into, not by as much as you all are used to us beating it by. So we can talk about what the surprise is, positive and negative for us, within the quarter, and run through that. If you think through, working kind of the way down the P&L, from an overall standpoint, outperformed, most of the revenue outperformance, you know, flowed through to the bottom line. So 7% margin on the bottom line in the quarter. And generally, you know, upside surprise there as well.
I think the key thing that we think about the business and the way that we've talked about the business for the last several years now is helpful just to remind folks, maybe ground them as we talk about Q1 and, and sort of the business going forward, is it's really helpful to think about the business in terms of, the growth in consumption of applications we've already won. So those are sort of existing workloads that we've already won from our customer base. And, you know, separately, to think about new business that we win in the quarter, and that new business can be from an additional workloads from existing customers or, or new logos, someone who's paying us for the first time. And the dynamic that we've seen-...
For about two years now, on the existing applications is slower growth, impacted mostly by the macro slowdown. And we talked about this a couple years ago. We were one of the first to point this out, to see what we see. And what really drives the equation there is we have a consumption model from a revenue recognition standpoint. So as people use and consume the service, that's what triggers the revenue recognition. And that underlying consumption is very closely tied to the underlying usage of the database. So if you think about the reads and writes, the transactions at the database layer level, not transactions like e-commerce transactions, but like the actual transactions at the database layer, and that's what drives the underlying consumption.
And so what we've seen is we've seen that grow more slowly, reflecting the macro environment. It's sort of not necessarily maybe macro in the way that you think of it. I can't ascribe a coefficient to GDP versus labor force participation versus, you know, inflation, et cetera, et cetera. But it's all related to the underlying actual business activity that's occurring in this broad set of workloads across closing in on 50,000 customers. It's fairly diversified across industries, geographies, use cases, et cetera, et cetera. So that's sort of the dynamic on the existing workloads.
And the growth of existing workloads, just given the size of the company now, as Serge said, sort of at a, you know, little under $1.7 billion last year, is the greatest factor in the short term of results. Over the long term, the most important thing is how many new workloads we win, given that we're, you know, low single-digit share in a, you know, $80+ billion market. But on the short-term basis, so in Q1, what we saw on the growth of existing workloads was we saw slower growth than we had expected. So we guided at the beginning of March. So obviously, we had February in the bag, and so February was natural, so no surprises, positive or negative in February, 'cause we kinda knew what that was when we guided.
What we've typically seen over these last few years is a seasonal uptick in the underlying usage and consumption in March and April. We did not see that rebound to the extent that we typically do. When you go back and you look on a year-over-year basis at the quarterly underlying usage, that read-write volume in Q1, it did grow at a slower rate than it did a year ago. So that's sort of what we're talking about in terms of sort of the incremental macro impact on existing workloads. As it relates to new workloads, despite sort of some of the challenging macro conditions over the last couple of years, we've been able to execute well against that front. We've been able to continue to win new business.
We have this, as Serge said, a very large market. We have a value proposition that's very strong and, and resonates within the market. And then, in general, the team has been able to pair that with really solid execution to sort of win despite more challenging environments. And so what we haven't talked about over the last couple of years are sort of this sales cycle elongation, deal slippage, and other things that some other companies have had to talk about. So in general, the new business side has gone very well for us.
We did talk about sort of one, you know, foot fault, self-inflicted miss, on Q1, such that the new business wasn't as strong in Q1, not having anything to do with market factors or anything, but we just, in the part of our annual planning cycle of assigning quotas and comp plans and territories and promotions and everything, we just were a couple of weeks slow on that, so just got a slow start to the quarter. But again, that's sort of a very specific thing as it relates to the new business side. So there's a bunch more to talk about, but maybe I'll just sort of stop there. But at least wanted to try and get into some of the things that will likely be on people's minds.
Great. And the other thing that seems like, I don't know, call it self-inflicted or, something that was unexpected, was the sales comp change at the beginning of last fiscal year. Can you just talk about what you guys did there and how you're tweaking that now?
Yeah, sure. Happy to talk about that. So we've been on a multi-year journey since we probably first started this in our fiscal 2021, which would have started in February of 2020, to move towards a more consumption-oriented model and less focused on maximizing upfront commitments. And there are really a couple of different theories around this, or a couple different drivers of the approach. But the core of it is, we have this very large market. We have a relatively small footprint of sales coverage relative to the opportunity.
And particularly when you think about negotiating a contract, you know, Jason, if you're about to launch an application, and my incentive as a sales representative is focused on maximizing your upfront commitment, you and I are gonna do a whole bunch of rounds where we're gonna negotiate and trade red lines, and I'm gonna try and get you to commit to a higher spend level than you're used to or want to, and you don't even know because it's a new application, and we'll go back and forth, and I'll waste all this time on this, and none of which will change what the application does, right? Or how much usage the application actually consumes. And so we wanted to be on this journey to sort of reduce that upfront friction.
It has the very specific and strategic purpose of synthetically expanding the capacity of our sales force, right? Because I can, you know, increase velocity, do more deals with less time spent per deal, that will allow us to capture kind of more market share. And in general, over the last several years, that's worked exactly as described and intended. So this past year, in fiscal 2024, we took sort of one of the final kind of key steps for that, which was we removed the incentive for sales representatives, for selling one-year commit deals, right? And so, again, the sort of final piece of that is, in that focus on workloads, and let's focus on workloads, and let's not worry about big commitment deals that could span multiple workloads, but let's go win as many workloads as we can.
And what you saw over the course of fiscal 2024 is we actually had a great year, in terms of the workloads. And what Jason's referring to is when we looked and started analyzing the data, and obviously, one of the challenges when you're getting data off of quarterly cohorts is you need to give them a little bit of time to sort of play out and see where the cohorts are growing. And those cohorts, you know, started off at sort of expected and grew at expected levels for the first couple of quarters. But what we're starting to see is some of those cohorts were growing. Their growth was slowing sooner than normal. They weren't exhibiting the typical sort of pattern or trajectory.
One of the things that we did or observed is that in removing that sort of final incentive around upfront commissions, again, if Jason and I are negotiating a deal, if I'm not looking for a commitment, I'm gonna lose a whole bunch of information or not have access to a whole bunch of information, or not gain a whole bunch of information that Jason's gonna disclose to me as we're going back and forth negotiating. I just kind of went in and grabbed a handful of workloads, right? And they wound up not having sort of what I'll call kind of like a representative portfolio of growth. But they...
The change that we made coming into this coming year was to be a little bit more prescriptive and not just ensure that you're looking only for, you know, workloads, and to recognize that not all workloads are created equal. But be a little bit more prescriptive to say, "Let's get some that look like, you know, A, and let's call these T-shirt sizes. You know, some mediums, some larges, and some extra larges," rather than just saying, you know, "A bunch of mediums or better." And these are minimum, you know, dollar spend thresholds that qualify as workloads. So that was a tweak that we made, based on, you know, what we were seeing.
We're constantly iterating, as we've done over these last, you know, 4 or 5 years, in terms of the go-to-market, but that's sort of specifically what was happening there.
Was that change made at the beginning of Q1 or the beginning of Q2?
We make all the comp plan changes at the beginning of the year.
Okay.
So they get implemented effectively in Q1.
Okay.
But those are. And part of the reason why you do it is, you know, comp plans are annual plans. You don't always have all the information that you need at the time, but you've got sort of your one bite a year to have a chance to go and do that. And so each year, we take our best guess and say: Okay, what are the tweaks that we wanna make this year, to try and improve the outcomes? Obviously, some of the incremental, you know, data after comp plans are rolled out, you get incremental data, right? And so, we continue to get data, and I think it sort of verified that this is a good change to make.
Is that one of the reasons why it took a little bit longer to kinda get rolling on the comp plans?
I wouldn't quite directly link them. To me, I think about it more in the context of fiscal 2024 was the first year that we had a workload-oriented model plan, you know, and all the associated territories and everything that goes with that. But we didn't have any data, which means there's really not much to debate, right? So this past fiscal year, you know, it was the first year that we got new data, and so as in the part of doing comp plans and everything for fiscal 2025, we actually had some data, you know, to debate and try and interpret and get into. And so it's not exactly related to what you're describing, but in general, that was an adjustment period, and we just, you know, took too long to do that.
Okay. And... I guess, what's your, what's your sense, I mean, what's your confidence level that the new comp plan is sort of Goldilocks?
Yeah. So maybe I'll say, I'll say a couple of things. Number one is, the reality is we won't know, right? We won't know for about a year because what Michael was saying, the, the workload cohorts we acquired last year was exceptionally strong in volume, and the original sort of growth characteristics were aligned with our expectations. So we gotta give it a little bit of time. Maybe one encouraging factor was that it was a very easy change to roll out to the field. You know, once we sort of have the sort of the new matrix in terms of the types of workloads that you need to go after, you know, the feedback that we got was: no questions, no concerns, not a tremendous amount of angst. More just like, "I get it.
Now let's go do it." So that's incrementally encouraging, but the reality is it will take us a couple of quarters.
Okay. And then one question that's come up has been: is it possible that the reason you're seeing kind of more of the kind of lower quality workloads, because a lot of the higher quality workloads have already been, you know, picked, the low-hanging fruit kind of?
Yes. So let me-
Before-
No, I appreciate the question. Quality is a word I tend not to use in this context, just 'cause it can mean so many different things to different people. But I think I understand, like, the point of the conversation or question. I don't find that plausible. And first of all, no one internally has offered up that as an excuse. And secondly, if I even were trying to think about it, it's just in a market that's $80+ billion, growing at, you know, $10-$12 billion a year, when you're a 2% market share player, that's just sort of not credible that, like, all the good ones have been picked, right?
Like, I think about the conversations we have with our customers, the workloads that we win every year, the pipeline, the conversations that we're having, and I just, I don't find that credible. I think it's more just sort of, you know, accidental, based on not having as much, you know,
Okay
... insight or intelligence into kind of the what was behind it, as a result of removing commitments.
Okay, and then two more sort of devil's advocate questions on like what could have gone wrong.
Sure.
One is sort of the sort of AI pause, you know, where customers are just, you know, taking a step back, and I think I know the answer to that, but I wanna pose that one. And then the second one is the idea that maybe there's a competitive issue with some of the hyperscalers, especially they are your number one competitor, I believe.
Yeah.
Correct. And, you know, customers are coalescing around their Gen AI development on those hyperscaler platforms. So maybe that's sort of creating some halo effect, and sucking in some market share. How do you think about this?
Yeah, let's take those two in order. So first, this AI distraction factor, let's call it that. So, we think that that's a plausible theory. And just the question is like: What's the magnitude, and how would it affect us? So let us first tell you what we see. We see tremendous amount of interest in AI in our customers, and it sort of spans the gamut from the developers all the way up to C-suite. So at the C-suite level, there's focus and pressure, sometimes even from the board, to come up with an AI strategy. And there's a tremendous amount of sort of confusion because the space is so rapidly moving, and it's sort of like a blank canvas, and it's really start to start painting on it.
So we see, based on the conversations that we're participating in, a meaningful amount of mental energy is going into this conundrum, if you will. Then on the developer side, we see them tinkering. We see them building proof of concepts. We see them wanting to learn the technology because it's new to them as well. Most of the pieces of the AI stack didn't exist a year ago. And then on top of that, they want to build applications to start to test in terms of how well do they work, and can they be scaled? Can they have the kind of ROI that they would need to actually productionalize? So in other words, we see focus on AI across the board. So could there be some amount of distraction and diversion of focus? Absolutely. But going back to...
So if that were the thing, the way it would impact us, it would impact our ability to win new business, meaning that developers and their bosses were not building as many, you know, traditional new applications, if you will, because they were focused experimenting with AI. And we don't think that, that would be a plausible reason to impact our new business. So what do I mean by that? Again, to what Michael and I were talking about, it's an enormous market. We have 2% share. Even in terms of incremental dollars, we're still in the mid-single digits. This AI distraction factor would need to be enormous. It would have to be something more like freezing or paralysis in order to shrink the available market so much to us that we couldn't reach our target.
So Dev said it on the call, he wouldn't accept that as a plausible explanation or excuse if somebody offered it to him for our new business performance. Our new business performance in Q1 is sort of we stubbed our own toe. We haven't changed our assumptions for the rest of the year, and even this AI distraction factor should not be distracting enough to impact our ability to win new business.
So before we go to the competition point, let me just try it back for people simplistically. If remember, I talked about two factors in terms of driving the business, right? Growth of existing workloads that we've already won, and then new business. So this, even the hypothetical of sort of an AI distraction factor has no bearing on the consumption of existing workloads, right? And then secondly, it's not been a factor in our new business to date, and you'd have to make some pretty Herculean assumptions. Doesn't mean that it can't be what other people are saying, or maybe other folks or business lines that aren't as where AI is more orthogonal to what they're doing, or they're a lower priority, but it's not what we've seen.
What is the split between the within AR, between the kind of growth in the existing workloads versus the new business? Have you disclosed that split?
We have not.
In a given quarter.
We have not. But conceptually, you know, if you kind of freeze your time at any given moment and then look at it before and after, the workloads that you have at the beginning of that period drive a significant portion of the growth in the near term, but those slow over time because they age, right? So you need to be adding incremental workloads, and those present a larger and larger percentage of growth the longer you go on.
Yeah, for those who haven't seen it, we walked through this and depicted a chart in our investor day back in May that sort of gives kind of an eight-quarter view and says, "Okay, here's what the installed base does." You know, if you're going to assume stable growth over the time period, ignore seasonality, and like those kinds of things, you'll see, you know, as the base matures, you'll see a slower and slower growth rate. Then you layer in workloads that you win, you know, in a given fiscal year, if we just constrain it to fiscal years, and then year two will have an impact. But the base, obviously, especially at our revenue, size, and scale, in the short term, is the biggest factor.
Yeah. Yeah, but I mean, in a given quarter, it's kind of most of the, vast majority of the business has to be coming from existing workloads. I would imagine-
Yes
right?
Right. But within that, remember, the prior years is a meaningful portion of that because those are the ones that are still, they're large enough to matter and still growing quickly.
That's still considered new business?
Well, if it's last year's, it's considered old business.
Okay.
But in the base, I think the difference, how old is it matters.
Okay.
If that makes sense.
Okay, and then the competitive question?
Yeah, so we've heard this one come up a lot. And sort of, maybe to paraphrase or repeat the question is something along the lines of, the cloud providers seem to be ahead of the game in terms of the AI pieces. They have more of them or all of them. So could they be winning all of those workloads and potentially even non-work, non-AI workloads because they're giving customers incremental sort of comfort to standardize on their platform, use just their out-of-the-box solutions, and therefore, you are seeing it in your new business, or rather, that's why your new business is hurting? Because stuff you would have gotten in the past, you're not seeing anymore because the cloud guys are boxing you out. Is that a fair sort of-
Yeah
summary? Got it. Yeah, so we sort of come at it at two ways. One is what we hear from customers when it comes to AI. So, well, sorry, before we go there, what you're effectively saying is cloud guys are bundling and, and boxing you out, and bundling has always been the strategy of the cloud players, right? They have the benefit of business walks into the door for them because they're there to buy infrastructure, and they've been using that for, you know, in the case of AWS, for a decade already, to sell them other stuff. Whether that other stuff is the database layer of any of the other 200+ services that they sell, that's been the bundling strategy. And obviously, we've done-
Reasonably well, we would argue over the years against that bundling strategy with our, you know, best-in-class, you know, platform-independent approach. Okay, fine, but maybe the world is different now. You know, maybe AI is the extra sort of secret sauce that makes bundling that much more competitive. And what we hear from customers makes us doubt that or not believe that. And the reason why I would say that is, number one is, by picking a platform, even the one that seems meaningfully ahead of everybody else right now, you're sort of betting that they are the, the technology winner in perpetuity, which, let's be honest, this AI cycle is what? 15, 18 months old.
You all are students of technology cycles, and you know that, you know, the early winner is sometimes the winner and sometimes not, and so by locking yourself in, you're foregoing that type of optionality. So the second piece associated with it is that you are increasing your dependence on that player. Like, you know, IT departments and CIOs know what the cost of a lock-in is when one vendor has more of their, more of your business. They have pricing power over you in the long term, and that's generally been the reason why they have multi-cloud strategies. One of the reasons why we've been successful is sort of an independent player. So now you're tripling on your platform lock-in if you start building your entire AI stack there.
Then the final one, and non-trivial one, cloud providers are offering proprietary solutions, not open source, and they're offering solutions that customers tell us are expensive. So ROI is one of the reasons why, you know, a lot of the development is happening at a lot of the AI action is happening at the infrastructure layer, but hasn't yet quite made it into large enterprise application layer, because some of the pieces of the stack are reasonably expensive. So you are foregoing the cost optionality if you're going with all the bells and whistles player, which, you know, the cloud providers is what they're bundling currently. Maybe we'll come at it in another way, which is, let's just say that that's true.
Let's just say that it is in fact true that we're being boxed out of more deals and seeing less new business because the cloud providers are keeping it more to themselves. We both partner and compete with them, and it stands to reason if they were more successful in competition, they would be less inclined to partner. And, and like historically, we've talked about how those relationships ebbs and flowed, but what we see currently is our cloud provider partnerships are stronger than they've ever been. And that's not just like executive to executive, that's all the way down the field.
You know, to kind of put it bluntly, if they were killing it with their first-party services, we would be calling them to partner on deals, and they wouldn't be picking up the phone, or they wouldn't be calling us, or they wouldn't be releasing marketing funds to go help us win incremental customers together, and we're not seeing that. So, like, we're deducing that this is not happening, but obviously we're very mindful of it and, and, and keeping an eye on it.
Good. I think maybe one more question before we wrap up the formal session. This is, it's kind of a couple of parts, so I'll try to make it quick, but the. I know you guys are using or thinking about using Gen AI to help the migration or help accelerate the migration of relational databases to MongoDB or relational workloads to MongoDB. My question is: Is there a risk that with Gen AI, it just becomes easier in general to switch out the database, and therefore, we've always known the database to be this super sticky space? Does Gen AI kind of fundamentally change the stickiness factor of databases?
So that's a good question. We are certainly working hard to reduce the stickiness of the relational database, right? And that's been one of the key inhibiting factors in our gaining market share, which is that, for all their challenges, databases and relational databases are exceptionally sticky, and taking these, like, in some cases, decades-old application and replatforming it, comes with a lot of labor and a lot of risk. But your question is fair. So if that sort of friction can reduce, why wouldn't all friction reduce effectively? And to which we would say, wouldn't that make it much more likely that the best database wins? And then you're back to performance, ability to scale, distributed data, developer friendliness, all the things that AI fundamentally doesn't change.
We think that's leveling of the playing field only accrues to the benefit of a 2% share player who happens to be best in class.
Okay, I guess we got 30 seconds left. Any questions from the audience? Last, last 30 seconds. Go ahead.
I'm happy. I'll repeat the question. The question is: How do large language models change our business or maybe our industry? I would say, generally speaking, we think of the benefit of large language models or AI to kinda come in three flavors for us. First, large language models will make developers more productive. Writing applications will become easier. We think that will result in more applications being built than they would be otherwise, so our entire industry is gonna grow, and the rising tide will help all boats, including our own. So that's bucket number one, and sometimes easy to forget 'cause it's relatively simple. The second one is, specific AI-powered applications will be built over time. We're seeing some early signs of that, although it's relatively early.
But back to some of our prior conversations, we expect that the database underlying those applications will need to be flexible, able to handle differentiated types of data, perform and distributed, and so forth, which means it's more likely that we win it versus the relational alternative. Then the final piece is, and maybe related to the first one, we believe that we can use large language models and other tools to make it much easier to replatform the existing estate of applications onto a modern platform like ours, which means that the part of the market that, you know, is accessible only as applications reach end of life, becomes more accessible in a shorter period of time. And that's also another thing that we're working on and try to benefit from. Yeah.
Oh, for sure.
All right, I think we'll cut the webcast now. Thank you guys for being here. Thank you, everybody, for staying late.
Thank you for having us.