We just realized, hey, this is not aligned. It does not make sense to slow roll it. We should just do it fast. Maybe it’s a little bit more disruptive at the very beginning of the year, but I think you get to something net positive, much, much faster in that world. And it’s — in some sense, easier than having one leg on each horse for an extended period of time.
Rohan Sivaram: And Michael, to your question around visibility, I’d say that when you look at the view of the world in an old scenario where customers are still committing ahead. Obviously, you have the visibility of RPO and CRPO, but in a world of consumption, it’s always trying to predict what the customer is going to do. It’s not ratable revenue. It’s always it follows a curve, you need to predict what your customer needs to do. And how do companies solve it, they’re essentially trying to do some predictive models by multiple factors, that predict that curve. In the new world, of course, what we are expecting and what we think is going to happen is the customer commitment is still going to be there, but they’re not going to be committing ahead of time.
They’re just going to be committing to your existing levels. So how do we get the predictability? As Jay alluded to, like — we get the predictability by understanding what are the use cases they’re going to drive. So the consumption transformation and all the work that we are going to do here will provide us, I’d say, added visibility into the work streams, which will help us predict revenue in a much better manner. And that’s one of the reasons why I called out in my prepared remarks that going forward, subscription revenue will probably be a better indicator of the health of the business.
Jay Kreps: And you can see exactly why it is, right? Like if we work with a customer, we’re trying to scope out every workload over the next two years and commit to lock that in. Of course, there’s some predictability like in the two years, we know what the minimum they have to spend is. But if we’re successful, they should be growing their consumption much faster than that curve anyway. And the effort of trying to really think of each application and how it’s going to be used and exact usage pattern and all the uncertainty that pushes on to the customer is actually just really high. And so taking that more out of the equation and just having customers really do it in a customer-driven way where customer locks in the right level of commitment based on what they know they’re doing, that’s effectively what all the peer companies have done.
It’s worked very successfully for them. And that allows you to then really orient the go-to-market organization around finding those use cases rather than getting people that kind of pre-commit to them.
Shane Xie: We’ll take our next question from Pinjalim Bora with JPMorgan, followed by Bank of America.
Pinjalim Bora: One clarification on the gaming company that you’re talking about. Can you help me understand was that company on committed contracts that move to open source, committed contract moved to Confluent self-managed? What was the transition exactly?
Jay Kreps: Yes. So, this was a set of applications running out of the cloud running on Confluent Cloud. That’s moving into their data center initially on open source Kafka.
Pinjalim Bora: Got it. Understood. Okay. That’s clear. One question on Flink in general. We have spoken to a lot of your partners. And it sounds like without Flink, and this is more of a high-level question, people are using Kafka streams on AWS with a Java-based application which is squaring into your Confluent clusters, right? That’s a compute that is not being captured by Confluent today. I’m assuming Flink kind of allows you to capture those compute. But then we’re also seeing AWS kind of coming up with their managed link service, right? I’m trying to understand Confluent being able — now that you have — you own the data persistence layer and the processing layer, as a whole, can you actually provide or offer something which is more than a point product Flink service that might be in the market?
Jay Kreps: Yes, we absolutely can, right? The challenge in the streaming world has been for a long time, customers are left kind of piecing together the parts their own to build a data streaming platform. So you get Kafka and then maybe you add on some Flink thing, you get somewhere else — and then the parts don’t work together, like the security models are different. The governance and catalogs of data are different, it’s all different. So it’s kind of left as an exercise to the reader to put those things together. That’s not the recipe for success, right? We think in this area, customers want to integrate it. So just in what we demonstrated at current already. I think we’ve taken an amazing steps. So unified security model across all of this, it handles all the connectivity.
Unified model of data. If you go and create a new stream in Kafka, it’s available there for you in Flink. You don’t have to go and recreate it in Flink again. The ability to actually just seamlessly use these two together and do it without having to preallocate anything. You can go issue a query against your Kafka cluster. It will scale up and run that processing and scale back down to 0 cost if it finishes. So already, that integration is leaps and bounds further than anything that’s been out there, I think, in the streaming world, integrating both the stream and the processing. And so if you think about, hey, what are our advantages in doing this, like stepping back to the question behind the question. I think the reason we feel like we have a lot of ability to succeed in this beyond any other competition.
First of all, Flink is this platform that has been incredibly successful that everybody wants. Second, we have the ability to integrate that stream with the processing. And if you think about the model for data systems, that’s just much easier. You could buy a database in parts where the query layer was separate from the storage engine, you can piece it together yourself, but that just doesn’t seem like the winning formula. And it’s obviously and very concretely a hassle today. And then finally, one of the things we’re seeing is that people are thinking about this processing and streaming together, and that’s moving upstream in organizations. And so, when we think about any competition in this space, most of it tends to come from different destination systems, but increasingly, the processing is kind of happening at the source and going out to all the destinations.
So structurally, those are the three factors that we think give Confluent a unique advantage in this. Our expertise in Flink and just the fact that we have the core people driving that forward, the ability to combine it with Kafka, the kind of de facto standard for the stream and really provide synergy between those things. And then structurally, the fact that the processing of data is moving upstream to the source where the stream is created, which allows us to bring together that combination at the right time.
Shane Xie: We’ll take our next question from Brad Sills with Bank of America, followed by William Blair.
Brad Sills: Wonderful. Thanks, Shane. I just wanted to ask a question here on this transition here, maybe one for you, Rohan. Should we think about this headwind, if you will, subsiding in the one-year anniversary from the change, or is there some other way to think about the timing of when that headwind kind of bottoms out, if you will?
Rohan Sivaram: Yes. Brad, that’s a great question. And as I shared in the prepared remarks, there are a couple of headwinds heading into 2024 which we spoke about a couple of customers, macro and the consumption transformation. And as you think about the shape for next year, we feel that the consumption transformation will have a bigger impact from an adjustment perspective in the first half of the year than the second half of the year. So that’s one data point. And I think Jay and I touched on it, exiting Q4 of next year, we’ll have a few tailwinds, and I’ll just reiterate a couple of them. The first one is the consumption transformation will be behind us. And what that means is how our customers want to consume confluence product and services will be very aligned with how we are going to market.
