MongoDB, Inc. (MDB) Earnings Call Transcript & Summary

All right. We're going to go ahead and get started. Thanks, everybody, for attending. Good afternoon. My name is Will Power. I cover cloud software for Baird. It's my great pleasure to have MongoDB here. MongoDB, as many of you all know, is really a leader in next-generation databases. From the company, we have Michael Gordon, who is the Chief Financial Officer, COO; and Serge Tanjga, who is the Senior Vice President of Finance. So thanks so much for being here.

Thanks for having us.

Thank you for having us.

Yes Michael, maybe just for the sake of the audience, just to help level set you can provide just a little bit of background on MongoDB and kind of your core target market and kind of how you differentiate yourself, and then we'll kind of jump into a fireside format.

Great. Again, thanks for having us. I know there are a number of journalist in the crowd who maybe don't follow all of us, so we'll take a little bit of a step back. So we at MongoDB make database software. So if you think about today, you hear these phrases about software eating the world or every company trying to become a technology company, really, that's all shorthand for the fact that today, companies are driving their competitive advantage from internally built software, right? That's how they're getting competitive advantage And at the heart of every one of those applications, whether it's a customer-facing application, a system of record, an internal application, those applications are built on a database and the database is at the core. And so that's what we provide. If you think about the need to have scalable, nimble, agile database to think about how much applications are changing, that's what really created the opportunity for us a number of years ago. And so we have the leading modern general-purpose database that we provide. It's an incredibly large market, $80-plus billion spent on databases in 2023. And growing at 11%, 12% a year. So the market is forecast to be, I think it's $138 billion over the next couple of years. And one of the reasons typically, I think, a large market like database maybe that should be more mature, maybe that should more grow in line with GDP. But at the core, the reason why it's not is because what I said at the beginning, it's so strategic and it's so foundational to how companies are driving competitive advantage. That's what really powering the growth when you think about the growth of new applications and companies being able to deliver better user experiences and differentiate themselves from their competition. So at a high level, that's where we compete.

Okay. That's great. Mike, of course, you just reported earnings in the last week. A number of moving pieces there. Perhaps just a maybe a quick overview, and we can jump into some of those in more detail.

Yes. So -- and I apologize, I seem to start losing my voice last night. So we'll get there as much until I expire. But we'll -- if you think about the quarter, we reported quarterly results last week, as Will mentioned. Top line growth was 22%. Atlas, which is our Database as a Service offering, which is now 70% of the revenue is growing at 32% year-over-year. We beat the top end of our guidance. And on the bottom line beat as well, bottom line margin, about 7%, operating margin in Q1. And generally, I think, operationally a fine quarter. It was not our best quarter. And those who follow it will know, but we'll probably get into that a little bit in detail. If I think about sort of some of the things that didn't go as well in the quarter, we saw slower growth from existing applications, and we'll talk about there a bunch of different subparts and reasons. And then for us, we had some self-inflicted operational missteps in the first quarter. January is our fiscal year-end. So this quarter was our first fiscal quarter. And we had a little slower start operationally from a new business standpoint. We mostly caught up but not quite by the end of the quarter. So that's sort of a onetime thing. And when we talk about macro and macro impacts on the business. I think it's really important to keep a few things in mind. One, when we think about the business, we think about the business in 2 pieces. We think about the new business that we're winning, the new workloads that we're bringing on board, whether they're new applications altogether whether they're migrations of existing technology. And for the last couple of years, despite the more challenging macroeconomic environment, we've been successfully able to win new workloads, and we're very happy with our progress on new workloads. The biggest driver of outcomes and results in the short period is what is the growth of the installed base, right? The existing workloads that we've already won. And that's where we see. We have a very tight linkage between the underlying usage of the application and what our customers pay. And what we've seen is slower underlying growth in the read/write activity at that. And that's when we kind of bake that all in, in Q1 and look out and give our updated revised fiscal '25 guide, that's where it's really factoring in. And despite the beat in Q1 had a lower outlook for the balance of the year.

So let's try to dig into that a little bit. I mean, you called out some of the self-inflicted issues, go to market, being a piece of that. Is there a way to kind of help frame how much that impacted the outlook versus what you're seeing on the macro front?

Yes. So maybe I'll try to do it and sort of do it in the pieces and where we quantify try to help people sort of understand the change in the guide. So I think the most thing for the investors who kind of closely followed the stock has been the fact that we lowered the guide for the full year. And we beat a little bit in Q1. So then if you think about the balance of the year, Q2, 3 and 4, we lowered the guide by roughly $40 million. So Michael mentioned a couple of puts and takes, but I'll sort of double-click on those and maybe add a couple more. But the first thing we said was if you look at our non-Atlas business, we lowered the guide for the year there because we're seeing less benefit from multiyear deals. So just as a brief aside, when we signed a multiyear deal for our Enterprise Advanced products or for some of our licensing deals, we take 606, we take the entire term license revenue component upfront. And so when you sign a multiyear deal that can be a significant benefit or a significant difficult compare a year later. Because of that dynamic, we're expecting to see a modest decline in EA business instead now we're seeing more like mid-single digits decline. So that's a part of the $40 million. That's a slight difference in terms of the rate of decline and then the rest is Atlas.

And so just on the non-Atlas piece, I think it's important to call it that is macro hesitancy to sign multiyear deals.

So somewhat consistent we would put other software providers have been recently saying. Then if you talk about Atlas, we -- there's sort of 3 components. We didn't quantify them or stack rank them, but let's try to sort of maybe explain at least conceptually how they work. First is the macro component. So we've seen slower usage than a year ago and therefore, slower growth of our assumption, and that's pretty uniform sort of across the base. And so we've had a macro slowdown 2 years ago. And some of the investors wanted to set a draw comparisons between that time period and this time period. What we see right now isn't quite as stark is what we saw 2 years ago when inflation was spiking when interest rates were going up, when startup funding dried up and people were bracing for an economic impact. We are seeing a slowdown in usage, but it's not sort of the same extent that we saw a year ago. But again, it applies to the entire base, so it matters. So that's bucket number one. Bucket number two is we called out specific workloads that we've acquired last year, so a subset of the base. They are important because they are still meaningful contributors to growth this year because they're still in the healthy growth phase. We've seen those growth more slowly than we've expected. So that's another element of contribution. And then the third is the turn on the Atlas side, that Michael was talking about the fact that we didn't quite deliver on our new business targets for Q1. We're not changing our new business outlook for the rest of the year, but that missing piece will stay with us sort of for the rest of the year. But then if you kind of put it together $40 million, a chunk of that goes to EA. The rest is Atlas. Maybe the best way to quantify it is like Atlas did $1.1 billion in revenue last year. So it gives you a sense of sort of the magnitude of total change between these 3 factors that we're talking about.

So maybe if I turn to unpack a little bit. So net new workloads at least through Q1 [indiscernible] but it sounds like you started to see some improvement, I guess, as you go into Q2, what's driving that? Or what kind of changed?

So you're going to say slightly differently and just following up on what Michael has said, Operationally, we were a little slower to start the year. So just in terms of just blocking and tackling that you got to do at the beginning of the year in terms of finalizing territory is finalizing more structure, issuing final quota numbers. We were just a little bit lower than in the quarter of 12 weeks that matters. And so because of that, we didn't quite get hit the ground running with our sales team the way that we would do in a normal year. And we actually mostly caught up by the end of the quarter. So in the end, we were behind, but we made it most of the way back. And so that gives us confidence that it's not a pipeline issue that it's not a competitive issue. Our win rates remain the same. It was an operational issue. But one that we know, a, we'll fix for next year, but in the near term, it just doesn't recur because now we're executing for the rest of the year.

Okay. And I think one of the other things you cited was go-to-market. And I guess there's -- the question I've gotten is, did you just emphasize quantity over quality. And where is that now in terms of getting that incentive structure right?

Yes. Maybe a few different things to think about. And I think this is specifically talking about the more recent workloads that we acquired, right, because newer workloads tend to grow more quickly, that's an important driver of our outlook for fiscal '25 as Serge sort of walked through. And so I would put this in the category of we continue to learn and iterate, and we've been in sort of a multiyear journey to consumption and to workload focus and reducing the importance of upfront commitments. So for those who've been around the story for a while, that will sound familiar and be very clear for those who are newer, what it basically means is that we've moved away over the last several years from an upfront large commitment even if it was just a 1-year commitment of how much are you going to spend and return maybe what discount am I going to get? And the reason why we moved away from that is because that negotiation of that discussion rarely impacts the underlying usage, right? And when you think about the Database as a Service offering, the way that we generate revenue from that is from the underlying reads and writes and the activity and the consumption on our platform. And so because we have this huge market, and we have relatively thin footprint coverage relative to the opportunity. There was a huge amount of wasted time with the rep has an incentive to try and get you to commit to $20,000 more or $50,000 more, none of which was going to change the actual underlying consumption of the subsequent year. And so we did is the final piece of that multiyear journey, which we started before COVID hit and everything else was removing the incentives for 1-year commitments that sales reps had. And so -- and instead really focusing them on workloads. So if you think about a rep's comp plan, simplistically. They're kind of 2 components, right? They're the net dollars of ARR that I drive and what's the new business that I'm driving. And then as we were continuing to iterate on this evolution, a quantity of workloads, right, that you had to acquire. And what we saw is that in reducing the commitment emphasis that sales reps wound up getting the velocity benefits that we wanted, right? We had a great year in fiscal '24, in terms of winning new workloads, right? We talked about that throughout. And so we got the kind of pay off or the benefit that we did. There's one thing that we learned, and like I said, I put this sort of in the category of unintended consequences that as a result of us now not having this long protracted time-consuming, inefficient conversation, I also missed some specific data about the application, right? So I kind of want a quantity of workloads. And our plan or our thought process had been, you'll win a portfolio of workloads and they'll look like a normal portfolio of workloads. What's happening is if you don't get some of that information, and the workload part of your comp plan is only focused on a uniform dollar threshold of like you need to be above the spend level in order to kind of qualify. There wasn't enough of a mix of the workloads. And so the tweak that we made to the current comp plan is same setup, same structure, but instead of just saying, okay, x number of workloads and many because plan is like we're going to specify the portfolio. You're going to get 2 that have a minimum spend of this and 3 that have a minimum spend of that, 2 is better than to sort of force the portfolio that was happening naturally and organically previously, but there's a chance that the focus and the evolution to workloads, you may be over-rotated too much to volume in part because you just didn't know because you weren't having that conversation and you couldn't say, "Hey, what's the -- what are the highest growing workloads going to be?" What are the biggest largest workloads going to be?

So where are you in that evolution that sales change...

We put that in place in Q1, and it was relatively straightforward. I've spent more time talking about it now than we probably had to explain it to a natural rep like they get it, they understand it. This is not going back on commits or rewinding the clock or anything. It's just saying, "Hey, have a little more intentionality around what the mix of the new workloads that you're winning look like.

Probably it's too early to tell whether you're starting to see some.

Yes, too early to tell, for sure.

Returns from that, okay?

Yes.

And I guess probably the single biggest question then is the change in Atlas consumption trends. And it sounds like you think just based on the broad-based nature, it ties to macro. But -- it sounds like it started in April and it was continued into May and you're just kind of assuming...

Yes, I'd say that differently. So if you rewind back to our prior call, which was in early March at the time we obviously had the February actual. So we knew what we were working on there. But the pattern of consumption we were expecting and that would be in line with prior years where the March and April are stronger than February. And they were a little bit stronger but not sort of to the extent expected. And once we sort of concluded that this is not just a week of tour aberration, but a more consistent sort of trend or lack of the seasonal recovery, we sort of dug into it and we really found it to be very broad-based across segments, across geographies, across customer 10 years. That plus sort of the year-over-year usage decline indicates to us that the phenomenon at play is macro.

[indiscernible] not a year-over-year usage decline but slower growth.

Thank you. And then -- and then in May, you saw trends consistent with Q1 trends and May is usually consistent with Q1. So that leads us to believe that we are in a slightly slower macro environment, but one that's been consistent now for several months, and that's what we're using to base our [ guide on ].

So are you assuming the normal seasonality from here?

Yes, normal seasonality from here, which would mean that the back half of Q2 is a little bit slower, back half Q3 is better, and we had the holiday slowdown in December.

Okay. And what provides the, I guess, conviction that it's macro as opposed to, I don't know, any kind of competitive share shift? I mean, it sounds like the new workloads suggest to you that in terms of win rate you're still stacking up well. But one of the questions I've gotten is hyperscalers because they're taking in so much more of this generative AI workload activity or they started to pick up some incremental share. What's kind of your reaction to kind of what you're seeing more broadly competitively that could help delay any concerns that might be out there on that front.

Yes. So I think it goes back to what Michael started, which is the best way to -- the best mental model for our business is to separate new and existing. So if we were seeing a competitive issue, you'll be on the new if we were seeing a saturation issue beyond the new if we are seeing the pricing or pick any sort of fundamental factor will affect the new side of the business. And we're not seeing it there. We're happy where our win rates are where they are, where they have been a very high. We were disappointed with our Q1 new business, but it's a function of a slow start as opposed to any fundamental change in the environment. And then on the existing side, the really underlying driver is the growth of the underlying application. And that's where we're seeing slightly slower usage, which we believe is a macro phenomenon.

Okay. Yes, I've got a number of other questions. But if there are any from the audience there, you do have structures where you can submit questions, and I can get those to the team here as well. Maybe kind of sticking with that kind of generative AI theme. I guess I'm kind of curious what you are seeing or what your thought process is around this idea that you look at the reacceleration from the hyperscalers, it feels like a lot of that is generated by generative AI applications. Is that kind of taking the oxygen out of the room in terms of spending in other areas? I mean, is that one of the areas that could be impacting you? Or is maybe that less of an impact? Any particular kind of viewpoint on where dollars are being allocated broadly from software?

Yes, I'll start and Michael will jump in as needed. So clearly, we're seeing evidence of GenAI spending in the hyperscaler numbers. What we hear from them and what we hear in the market is that it's happening sort of at the infrastructure layer. So a lot of the money goes into GPUs model trading so the sort of activities that predate effectively the creation of apps. And that will be sort of the next layer up the layer where we play and the layer where we would expect to see benefit sort of in the medium term. So we can get back to like what that means and what it looks like. But sort of sticking around this sort of near-term narrative. What we heard from investors is, okay, well that money and it's significant must be coming from somewhere. And where do you guys think it's coming from? Is it coming from software? Is it coming from in-house app development, where exactly is it coming from? Obviously, difficult for us to answer from the perspective of, look, we have small market share. So our visibility would be less than if we were the cloud provider themselves. Total logically, the money is coming from somewhere. And we anecdotally see that customers are playing with AI that developer teams are building proof-of-concept applications that the C-suite is engaged to figure out what exactly AI means for them, what is the strategy and how they -- how most quickly do they execute in a cost-efficient way. So there's definitely an element of AI distraction that is sort of impacting the market and maybe softening all the dollars. Now how that impacts or doesn't impact us. Again, in the new versus existing paradigm, it would impact the new it would impact our ability to go win workloads that are currently being built because there's fewer of them being built while people are paused their regular software development in order to focus on AI. We don't think that this magnitude of switch is such that it should impact our ability to generate new business or as our CEO, they would say, like he would not accept that as an acceptable excuse, only because our market share is so low to Michael's point, the market is enormous. You would need to have some sort of AI winter fully freeze regular sort of ongoing business and software development for it to really impact us given our starting point. And so yes, anecdotally and conceptually, it makes sense that it will be happening. We don't think it's happening at a scale that is impacting our new business. And just to close the loop, it shouldn't be or doesn't have anything to do with our existing business.

Just to your hyperscaler point, Will. I mean I think, a, a huge amount of that is on the training side. There's very little that's taking place on the inference side, certainly from everything we've seen and heard, not just from our own business, but from talking to everyone else, talking to customers, things like that. And then I think also people try to do these read across where they look at the hyperscalers and try and say, what can we learn from that? And I think one of the things that's important to understand for the hyperscalers is a year ago, they were under a fair amount of cost pressure, optimization headwinds and all to those kinds of things that aren't our dynamic and even though that we recognize revenue on the basis of consumption I think sometimes people mistake that revenue recognition and make it equivalent to our business model. The business models are a little different. And so they went through a series of optimizations that we didn't go through that they're now lapping now. And so they're sort of headwind from last year becomes a tailwind for them this year, but that's just not our dynamic. And so I think that also helps like put things in context of folks.

Yes. No, that's helpful. I guess maybe kind of stick with that being a little bit. I mean, you're coming off of local New York here just a few weeks ago. It seemed like a lot of enthusiasm there. Any kind of green shoots you'd point to with respect to AI application development? I mean that's kind of what we're kind of waiting for, right, is those applications that you can help underpin effectively moving forward, and maybe a lot of those lines with Vector Search, the integration you have there with the broader platform, what are you kind of seeing on that front?

Yes. I'll comment in a couple of different ways. But the one thing I would say is we are not waiting for in the sense that there's plenty of opportunity to go win new business. unless -- until sort of the AI wave actually washes out. So just -- we think of that as an incremental vector of growth out of, already plenty of other incremental vectors of growth. But where we're seeing interest is sort of what we've seen interest primarily and where we're seeing it for the past year is a tremendous amount of innovation by the start-up community. So we have pages and pages of referenced use cases of startups doing very cool things on MongoDB using all sorts of varieties of data from video to voice, to PDFs to combination of all of them to build interesting use cases, but they tend to be new. They need to build the product, then they need to find the product market fit and all of that. But it gives us a lot of confidence in our positioning in the market to see just the amount of innovation that's happening on the kind of start-up and in the market for us. When it comes to enterprises, they are very interested back to prior conversations, the conversation really spans from the C-suite all the way down to the developer and they're thinker. They're building proof of concepts. They're excited about when those come together. The question is, a, they're hard to build; b, they are sometimes expensive to build because some of the pieces of the AI stack are quite expensive still. And then finally, they need to have confidence that they can deliver the user experience that's necessary to generate the ROI on that investment while solving things like security, like IP and on and on and on. So you see a lot of experimentation. We hear of exceptionally cool outcomes, which gives us confidence that they will be scaling some of those applications into production as the time goes on, but it will take a little bit of time.

Okay. I've got one question, Serge. I think it's kind of a clarification asking if new business trends have improved recently, and I think that was coming back to the net new workload comment.

Yes, that's right. So we started off slow in Q1 and it has been better since then.

Yes. I would just clarify, that's not a comment on the market. That's a comment on. We got a delayed start to the year and then we pushed well through the quarter and almost caught up, not somehow we are seeing some change in the market dynamics or anything like that. That's more of an internal execution, operational issue.

Right. Okay. Maybe just to circle back to MongoDB event, local here. Anything else that really stood out for customers or customers were particularly excited about or you came out thinking, this is something -- going to start to get traction on. I mean, what's kind of the next product, I guess, that you're most -- maybe most excited about our product or 2, that you think can start to generate more usage here over the next 12 to 18 months?

It actually wasn't the product per se that we got most or maybe most relative to expectations, customer feedback. But it's this program that we announced called MongoDB AI Applications Program where we've announced effectively a series of partnerships. That makes it easier for customers to get started with AI. So we've curated a set of partners in basically every subsector that are needed in the AI stack. So whether that is the hyperscales themselves are participating, whether that is model providers, model hosting companies, embeddings management, orchestration frameworks, including services providers as well and sort of give a list to customers of like [ Cohere ] partners that we believe we, MongoDB, believe are innovators in the space. And if you're looking at how you're going to piece together your AI application, you should start with this list. And the feedback from customers was positive because they need help to cut through the noise. The noise is significant and growing. And every week, if not every day, there's a new press release and a new leader in every one of those quadrants. So they need assurance from a credible third party of like where should they focus their efforts. And also assurance that over time, as the market develops, we will be there to help them. And we'll do more map. We'll do things like reference architectures, we'll do things like packaging or sort of co-selling arrangements. But the feedback was positive and sort of validating 2 ideas. One that we play, a, critical part of the stack and b, that customers really need help to figure out how to put it all together. And then the second thing I would say is, which we've been talking about for a while and talked actually at the investor session about is, it's also AI related, which is that we see great opportunity to use AI, to make it easier to migrate off of relational databases and on to MongoDB. We call that app modernization factory. We talked about a few pilots that we've done to start the year and several more coming in the back half. And the first results are very encouraging in that AI, meaning large language models make it much easier to rewrite the code of a Legacy application a significantly easier and less risky to build testing to make sure that the new MongoDB application performs the same function effectively as the relational migration -- the relational application. So those 2 things together as well as the feedback from the first pilot gives us a sense that the opportunity is increasingly feeling real. It's going to take time to execute because, again, we do one customer at a time. But those are the things that in and around are local and sort of fortified by customer feedback gives us a sort of incremental confidence of those vectors of growth going forward.

Yes. I would just underscore the relational migration piece, app modernization, whatever you want to call it, very early, but over the long tail, should help, and it's a great example of sort of using AI, not necessarily in an AI application. But how that's increasingly important and certainly helping us or as a potential to help us over time. The other perspective that I would just add is one of the beautiful things about MongoDB from my perspective is we have this very small share in this incredibly large market. So we don't need some new product to be some sort of like big unlock or access some adjacent TAM or power future growth opportunities as boring as it is, a lot of it is just about executing into the opportunity that we have against us, in front of us. That said, we are investing against the broader vision and a broader road map and excited about what that holds in the future.

I think one of the other broader questions I get is trying to understand why you are well positioned to help customers with generative AI and build generative AI applications versus some of the other platforms out there, Data Warehouse approach, Data Lakehouse approach, Snowflake, Databricks or even the hyperscales. What's kind of your answer to that? I mean, obviously, you've got great developer presence, a great set of capabilities. How do you kind of set yourself apart to be that kind of go-to partner?

Right. So if you think about building an application [ So our person has ] developed. So not an analyst or an MLOps engineer or a data scientist but just your developer. And so if they're building an application, they are going to require a number of components, but ultimately, what an app does is the right data. And for that, you need a persistence layer, an OTP persistence layer and that's the core space that we're performing. That's not a function that a Data Warehouse or a Lakehouse or any other sort of repository already created data serves to perform. And every application includes AI application needs a database, and that's not going to change going forward. So that sort of defines our competitive set for that portion of the AI stack. Now within that, what's becoming increasingly obvious to us, not just because it makes intellectual sense to us, but also because we're hearing it from both customers and partners is the importance for that OTP layer, the database layer to be flexible, to be performing, to be fully distributed and to be able to work with heterogeneous data. And our sort of foundational architecture advantage i.e., the document model is perfectly suited for exactly that type of use case. And that's different than relational to rows and columns. It's different, frankly, for effectively all the other no-SQL solutions that don't have that flexibility, versatility and heterogeneity data. And that is going to be needed as you think about the exciting AI applications that we think are going to power our lives, whatever, 5, 10 years from now. And again, essentially it makes sense. It's made sense to us since just the beginning of the AI conversation over a year ago, but that's what we're hearing brings customers into conversation with us, and that's what brings partners to the table if they're want to partner with us.

And then just more broadly I know we're up on time. Just more broadly, I think our view is that applications will get smarter over time, whether they're actually AI-specific applications or other applications that just become sort of more intelligent AI-powered, whatever you want to call that. And a couple of key trends that are important there are: one, that's the domain of developers, developers creating those applications. So we think more and more responsibility is going to shift to developer where we're incredibly well positioned. And secondly, it's leveraging that operational data that's being created in the application. It's not some stored after the fact and some Data Warehouse or some Data Lake. This is the operational data being created in the application itself.

Yes. Now that makes a lot of sense. All right. Well, thank you all. Those are great insights. Really appreciate it. Please join me in thanking Mongo for their comments here.