pingingz
Confluent laid an egg-can it turn into a tasty omelette or must it be discarded?
This earnings season has been a mixed bag. There have been some notable winners reflecting stronger than anticipated demand-Amazon (AMZN), Square (SQ), Shopify (SHOP), Arista (ANET), Affirm (AFRM) Datadog (DDOG) and Microsoft (MSFT) are prominent in that group. And then there have been some companies whose results or mainly guidance have reflected the reality of a difficult IT spending environment or some company specific issue. I have to include Bill.com (BILL) as part of this segment, but in addition, Google (GOOG), Paycom (PAYC), Fortinet (FTNT), Procore (PCOR), Trade Desk (TTD), Veeva (VEEV) and Atlassian (TEAM) are all part of this less favored group.
What should investors do now? Pile into the winners or try to see if there are some hidden gems amongst the rubble. The latter strategy is not for everyone. Some companies on the above list are sick, some have a mild case of the flu and some are allergic to macro headwinds.
Sad to say, my guess, for what it is worth, is that macro headwinds are more starting then in a penultimate phase. This is not an article about the chances of a recession and how that will impact IT demand. I personally believe a recession is coming, and have had that belief for some time. And a recession will and is already creating demand headwinds in the IT space. IT demand has always had both a growth and a cyclical component. Companies are reluctant to invest in infrastructure of just about any kind-IT or otherwise-when their own businesses are seeing faltering demand. And some of that is already visible.
On the other hand, we have recently had reaffirmation about the links between long-term rates and valuation-both positive in 2021 and a part of 2023 and negative in 2022 and a part of 2023. There are many articles and viewpoints about the implications of last week’s varied metrics and supply factors in the bond market to account for the significant rate pullback. And there will be additional articles after today’s benign inflation print. This is not an article that attempts to handicap rates, other than observing that shares of Confluent (NASDAQ:CFLT), along with many other high growth IT companies will see their valuation correlate inversely with the direction of long term rates.
This article attempts to lay out the case to buy shares of Confluent at the current price. Patience is almost certainly going to be required. And the road to recovery is likely to have some bumps and twists. Many institutions who invest in the IT space wait for these kinds of opportunities. They simply never will otherwise be able to establish a significant position at a reasonable valuation.
Sometimes commentators of various stripes including institutional analysts take the obvious path of recommending what is working and reporting strong results. I confess I often find myself doing that. And other commentators seem to feel that they are shedding light on a company whose shares are down by ~40% in a single day with a new sell recommendation.
There are companies I have recommended consistently and they make up the preponderance of Ticker Target’s high growth portfolio-for example I have recommended and owned Microsoft shares for more than 6 years on a consistent basis. But sometimes I feel that a high growth portfolio needs to include companies with the opportunity for sustained hyper growth that are not currently achieving their long-term potential-a bit of leavening so as to speak. And Confluent is one such company.
In developing a recommended high growth portfolio for Ticker Target subscribers I attempt to include various kinds of software/technology stocks. I look to have diversity by function; i.e. I have cybersecurity, observability, workflow management, infrastructure and ecommerce recommendations. Some are hard to characterize. Many recommendations are representative of the space and are of companies that are currently doing well and which have strong visibility.
Confluent is not one of those-but I think it is worth owning-basically because it is the leader in what it does, and what it does is a key component in a GenAI world which is just now emerging. Many developers have opined that they really rely heavily on Kafka as a key tool. And Confluent’s version of Kafka is made for the large enterprise. Don’t buy Confluent as an “AI company.” Do buy the shares because its versions of Apache Kafka and Apache Flink are tools that make the Gen AI world work better, in terms of efficiency and time to benefit. Data streaming-the core Confluent technology and what is called data processing (Flink) provide functionality that is important now and will be more important over the years simply to make GenAI work properly.
I am a software analyst and a growth stock investor. I need to make clear that this article does not focus on Confluent’s valuation. I doubt that there is a sentient investor who would choose to buy Confluent shares for the 22% growth it has forecast for 2024. And while there is an expectation that Confluent will be profitable and generate cash next year, certainly no one is going to buy the shares because the non-GAAP EPS is expected to be marginally positive. My estimate for Confluent’s EV/S over the next 4 quarters is still greater than 8X, and I doubt that the free cash flow margin will be anything above the low single digits.
I am not recommending the shares because of that valuation, but because I think the company, especially with the acquisition of Flink, has an opportunity to become a large and profitable hyper growth business. It can be thought of as the plumbing of and application and network structure-not visible to some, but absolutely essential for comfortable functionality. When that is actually visible, the shares will have long since appreciated to a different valuation tier.
Investing in hyper-growth stocks is really not about creating voluminous models and spread sheets-although I do that as well, but about evaluating a market opportunity, a product strategy, go-to market plans and the quality of management. Stream processing as it is called, is a big deal in the modern software world, Kafka is essentially the defacto standard for stream processing, Confluent Kafka is enterprise grade, while the open source distro is not, and I believe it will be very difficult to create efficient AI applications without the use of Confluent’s technologies. That is why the shares should be bought-and not because of 2024 estimates.
Confluent: What happened?
I last wrote about Confluent shares for Seeking Alpha in April of this year. The shares are down 27% from that time. Of more significance, the shares are down 54% from their high set in July, and down by 38% since just before the release of quarterly earnings on 11/2/23. I have never owned Confluent shares-more a matter of luck then design-but I am planning on taking the plunge in the coming weeks.
The company’s most recent quarter was a beat compared to prior expectations. The company was non-GAAP EPS positive for the first time. The problem was that the company reduced its guidance for revenue. It is now forecasting revenues of $205 million for the current quarter, it had been forecasting revenues of $212 mil. It did not reduce earnings guidance for Q4-the non-GAAP EPS forecast has been left unchanged after adjusting for the Q3 beat. The company has focused on improving its margins; it hasn’t quite gotten to consistent non-GAAP profitability and while the cash burn hasn’t ended, it has been reduced substantially.
The company, while hitting its overall revenue target was below its internal expectations for cloud revenue-although the miss was marginal. With regards to the $7 million forecast reduction for this current quarter, about half of that was the result of a couple of company specific issues with two large customers. In one case an on-line gaming company moved its workloads back to their own data center. This was not particularly a Confluent issue, but it had a visible impact on revenue. In another case, one of the company’s other large customers has ramped more slowly than anticipated because it is being acquired. These things do happen-although a company moving back from the cloud to on-prem is quite unusual.
Still with just over $200 million in current quarterly revenue, and a difficult environment for IT sales and consumption growth, a couple of unanticipated issues with larger users will have a visible impact.
What was more of an issue, at least to me were some comments during the conference call regarding overall demand issues. The company called out continuing macro pressure, including the ongoing conflict in the Middle East, where Israel is a top 10 country in terms of Confluent’s business, and the possible U.S. government shutdown (now adverted for some time), both of which add uncertainty and disruption in particular segments. Specifically, Confluent has seen slower organic consumption growth resulting from a slower rate of new use case additions in some part of its customer base. It didn’t specifically call out lower growth in consumption from its installed base of cloud users.
Confluent’s sales issue with Israel is unlike other IT vendors, some domiciled in that country. Apparently Confluent’s Israeli customers have taken a step back when it comes to developing new use cases based on Confluent Kafka. It is unusual that Israel, despite a robust IT sector, is a top ten user for a company, but that is, or has been the case for Confluent.
Outside of Israel, overall, macro headwinds are reducing consumption growth as well-somewhat similar to what was seen several quarters ago by other companies such as Datadog, Snowflake and Microsoft although the issue here is more the development of new apps rather that the consumption of compute and storage of existing apps. Just how much the possible government shutdown impacted consumption is fairly questionable.
I have written many times, and about many companies with regards to consumption and cloud optimization. Confluent, perhaps because of its size, had heretofore avoided any impact from cloud optimization and consumption headwinds. But the company exists in a world in which macro exigencies are leading many enterprises to examine both new projects and try to reduce expenditures on current apps. I think the share price reaction to what has essentially been an industry wide phenomenon is more than a bit overdone.
Beyond that, the company is making a change to its go-to-market motion. The company is moving to orient its cloud business segment to a model centered entirely on consumption. This has been a planned activity, but will be accelerated heading into 2024. This will mainly impact the compensation of sales people; the company has a pricing model that is mainly consumption based for its cloud customers. Historically, sales people have been incented to get user commitments ahead of usage; users are balking at that paradigm, and instead the company will start to pay for incremental consumption and new logo acquisition.
The net of all these headwinds, including the sales compensation transition, was to reduce expected revenue growth in 2024 to 22% from a previous estimate of 28%. I don’t think that a 22% revenue growth number by itself would be considered as a disaster in the current environment. But the issue is-that’s a forecast and many investors and analysts have concerns that the forecast might not be achieved, or that the sales compensation transition will have impacts longer than expected, or that the environment will continue to place a drag on revenue growth. I certainly don’t have anything like second sight when it comes to trying to handicap Confluent’s expected revenue growth in what is a tough environment for IT sales. When queried as to the basis for the 22% revenue growth forecast, the CFO offered the following.
Great question, Sanjit. Yes, when thinking about our 2024 guide, I kind of put it into three buckets. Starting off with, we called out the two large customers that had impact, although customer specific implications, but that will have an impact into Q4 as well as 2024. The second category is around the macro, which, I’d say, a combination of geopolitical exposure to Israel, coupled with the slowdown in the new cases that we’ve been seeing. And the third category is the consumption transformation. And as Jay just called out, any time you go through a transformation like this, there’s going to be this adjustment factor. And we are trying to prudently bake in that impact of that into our guidance.
So, taking a step back, these are the three drivers that impacted our guidance. But when you think about the first half versus second half, we expect like the second half to be better off slightly than the first half, clearly because the consumption transformation will have, I would say, a larger impact in the first half of the year. I mean, I also want to call out, Sanjit, that as we are exiting 2024, there will be a decent amount of tailwinds. First of all, the consumption transformation, which we expect will be behind us and which will reduce the friction between our go-to-market teams as well as how our customers want to buy our products. Second, I mean, we’ll have a decent amount of product tailwind behind us. We’ll have Flink, we’ll have GA about 6 months in and a couple of other unlocks from the data streaming platform perspective, coupled with FedRAMP and AI. So we feel that exiting 2024, we feel pretty good with where we are and just in general, from a long-term perspective.
I don’t think that any of the above is somehow contrived, or unreasonable, or self-serving. But as is the case with most forecasts which deal with the future, there is no way of validating these kinds of expectations. Certainly, the two specific customers who were called out, will only have a finite impact on demand. And it is just as likely that over the course of a year, the company will locate and close some unusually large, and unforecasted opportunities.
But how long and to what magnitude the decline in use case growth might last is really unknown and speculative. As I will try to explain, my view is that the macro environment is wielding a scythe through almost all phases of the infrastructure software stack. Until now, Confluent had avoided the broader industry issues until now. But those issues, compounded by a couple of company specific customer demand problems and some go-to-market friction eventually caught up with Confluent. This does not mean the wheels are coming off the company or that the growth thesis has been invalidated.
It ought to be obvious at this point that many, if not all software companies are going through demand issues. Even a company such as the Trade Desk, the leading ad tech vendor, has not been able to escape macro headwinds. The fact that Confluent is forecasting lower growth in 2024 doesn’t indicate some deep seated issue; it reflects the fact that overall market dynamics have eventually caught up with the company and given its need to transition to a sales incentive model entirely based on Kafka Cloud for incenting its go-to-market teams, this has led to an inevitable growth issues although the end result will better align sales incentives to the company’s overall sales strategies.
Indeed, compared to other companies I have seen reduce guidance, the company went out of its way to suggest that demand tailwinds would return by the end of next year as the quote above strongly suggests. The next couple of sections will deal with the product strategy that Confluent has embraced, and why, in my opinion, this is a strategy that will produce hyper growth, with solutions that are differentiated.
What does Confluent do?
For some readers, no doubt, what Confluent provides may seem a bit obscure and not of existential importance. But in the software development world, what Confluent provides is of great significance and growing in importance. I am not going to try to do a deep dive into the all of the potential benefits of using Confluent Kafka. I also want to make clear that my knowledge of both Kafka and Flink is not the deepest. Not all readers will feel the need to try to understand what this company’s solutions do, and some readers may feel that they don’t need any kind of primer on Confluent functionality in order to make a decision with regards to investing or not investing in the shares.
The company’s core technology is that of what is called data streaming. Historically, queries have been made using data at rest. This has many limitations. One example would be creating an application that provides a seamless experience for the customers of ride hailing companies such as Uber and Lyft. Those apps would not work with data at rest; there is dynamic information about demand, vehicle location, traffic data and driver availability that needs to be incorporated and continuously changed to provide the user experience that we now all expect when we go to those apps.
One fairly typical use case is that of real-time inventory management. Michelin (MCHA.F) and Instacart (CART) both use Confluent as an inventory management optimization engine. I confess I do not and never have used Instacart-it isn’t really meant for my generation or age bracket. But essentially, the Instacart experience relies on its customers having access to real time data so that they can buy what they want, where they want and from whom they want. It just wouldn’t work without access to many different data streams all being accessed simultaneously.
An important vertical for Confluent is that of financial services. The use of Confluent cuts across many functional areas. It is used for compliance, fraud prevention, risk mitigation, insurance etc. I have linked here to brief reviews of 68 Confluent cases from a service called CASESTUDIES.COM. For readers who want to get an in-depth sense of the value that the use of Confluent can bring to users this might be an interesting read.
Confluent is based on the open source distro called Apache Kafka. It has nothing to do with the Czech author of that name. Kafka was initially developed at LinkedIn and then moved to open source about a dozen years ago. Two of the developers of Kafka, Jay Kreps and Jun Rao were the founders of Confluent and are still leaders of the company. There are many differences between the open source version of Kafka and the version that Confluent sells. I have linked here to a 3rd party evaluation of the differences. To summarize, Confluent has a much more user friendly interface that is designed for the devilment of event-driven microservices, IotT applications and other complex, mission critical applications.
Kafka itself is wildly popular and is used by 80% of Fortune 100 enterprises. At this point, Confluent has about 1200 significant users. These users are paying for Confluent, rather than using the free Kafka service because of the features and support Confluent offers.
The availability of additional features and tools that come with Confluent is one of the main distinctions between Confluent Kafka and Apache Kafka. Confluent Schema Registry, Confluent Control Center, and Confluent Connect are some of these features and together, they offer data governance, monitoring, and integration capabilities. Considering these additional features, Confluent Kafka, seems likely to be more suitable for use cases at the enterprise level
The degree of support provided is another significant variation. Kafka is supported commercially by Confluent, but Apache Kafka is an open source project that depends on user support. For companies that need high availability, dependability, and prompt issue resolution, this support is especially crucial.
As can be seen in looking at the use cases to which I linked above, when enterprises slow down the development of new features/functions for their own customers, that ripples through to demand for Confluent. There is nothing mysterious or particularly surprising in that observation; the use cases all are related to helping companies improve their operations, and inevitably have an ROI, but in a recession those kind of improvement tend to be delayed or downsized and that is basically what is happening to the growth of Confluent demand.
I am not going to try to evaluate at this point the probability of a recession or the hardness of a landing. In the case of Confluent, the recession has come, and its expectations for growth are being constrained by that factor. This is no different than results/forecasts that were reported by the Trade Desk, Bill.com Atlassian and Veeva. There are exceptions, to be sure-Monday.com for a variety of reasons is not at down with “recession flu” but seems to bursting with good health.
Many observers, including this writer expect that generative AI will create a strong demand tailwind for much of enterprise software. Confluent unveiled its particular Generative AI capability at the end of Sept. Steam processing is going to be a foundational technology for just about any kind of Generative AI. Here is a comment on the subject from Confluent founder and CEO, Jay Kreps,
For the past decade, AI heavily relied on historical data, integrated with slow, batch-based point-to-point pipelines that rendered data stale and inconsistent by the time it arrived. That’s no longer adequate for the real-time AI use cases today’s businesses are trying to launch, like predictive fraud detection, generative AI travel assistants, or personalized recommendations. Compounding the problem are issues with poor data governance and scalability. As a result, the pace of AI advancements is stifled as developers are constantly tackling issues with out-of-date results and untrustworthy AI hallucinations. This isn’t just a technical hurdle; it’s a roadblock to AI innovation.
I think it is reasonable to believe that the vast preponderance of AI applications in the enterprise will need stream processing to produce anticipated results. The adoption of Confluent Kafka has hit a speed bump-but is a speed bump rather than a detour or a sustained sales execution or competitive issue. At some point, just about all enterprise software companies outside of cybersecurity are or have been dealing with the same speed bump. It has to do with macro conditions and not the ROI or necessity of most modern software applications.
What is Flink – Why is it important for the future of Confluent?
At the start of this year, Confluent bought a company called Immerok. Immerok is a leading contributor to the Apache Flink project. It’s a big deal for Confluent, far beyond the scope of the financial consideration and the immediate revenue contribution. Just to be clear, at the time of the acquisition, Immerok had not yet started to offer a fully managed Flink offering that was compatible with the Confluent Cloud. There are likely to be substantial revenue/cross selling opportunities. Some observers actually believe that Flink has a greater total opportunity than Confluent Kafka.
Flink is designed to process massive amounts of streaming data from many different sources such as message queues, socket streams (this is a stream of data from A TCP network server) and files. There are a fair number of nuances; Flink started as a stream processor that has added batch processing. It can be viewed as complementary to data lake vendors such as Snowflake. One way of looking at Flink is that it is the software connection that makes sure that raw data gets to where it needs to be at the right time. Inputs have to be handled quickly and safely in order to ensure that an application provides a user experience of a level that is necessary for an application to provide significant benefits for a given enterprise.
Flink and Kafka (Confluent) are different yet very complementary. They rely on each other in order to create the best user experiences and functionality. Think of it this way. If an application is built on Kafka, i.e. the use of streaming data, but doesn’t use stream processing, its performance will be less than optimal. Modern applications need both data streaming, but also stream processing. No matter how efficient Kafka is, without stream processing, an application developed using that tool will sit idle part of the time waiting for data to process in the application or analytic process. Flink and Confluent Kafka are far more than the sum of the parts.
Flink is going to be generally available integrated with Confluent by the end of the 1st half of next year. The initial announcement of the new platform was made on 9/26 as is shown in the above link. There will inevitably be a gestation period for the new offering. But my view is that with the combined offering, the expected CAGR for Confluent will be substantially greater than when it simply had its version of Kafka to sell. I expect this new offering will have a peripheral positive impact on growth in 2024 but will be a significant component of the growth story in 2025.
I don’t want to suggest I have some particular crystal ball regarding how fast Confluent will be able to grow with the combined product. If revenue growth for Confluent in 2024 winds up at 22%, then a minimum growth expectation for 2025 ought to be greater than 30%, the same growth rate as the company is forecasting for the full year of 2023. Can revenue growth be 40% in 2025?-I certainly could make a case that it might be given the less stressed comparison, new products, the absence of an impact from two clients and the tailwind from a remediated go-to-market paradigm with appropriate incentives.
There will be many different opportunities for cross selling including SKUs for connection, processing, governing and sharing. In addition to cross-selling, the combination will help Confluent become a significant component of the plumbing of Gen AI apps. Of course that will take time, and many investors are looking for instant gratification when it comes to AI tailwinds, but in considering a 3 year CAGR, the combination of Flink, Confluent Kafka and Gen AI will be a formidable demand driver, in my opinion.
Retooling incentives for the Confluent salesforce – A necessary one-time headwind that will reverse direction to a tailwind
The Confluent Cloud has a hybrid pricing model. There is a monthly charge per use case and charges for usage and for data ingestion as well. In addition there are charged for stored data and charges for data ingress and egress. The company wants to incent salespeople to focus on opportunities that maximize cloud subscription revenue rather than bookings or committed spend.
The company had previously intended to phase in the changed incentive plan over 3 years; it is now being adopted at the start of 2024. It makes sense to pull the Band-Aid off the festering wound; given that demand issues overall are going to be a factor in 2024 revenue growth, making a necessary change at one fell swoop makes the most sense.
The company believes that these changes will create a 2024 headwind for margins of 200-300 bps because more commission expense will be recognized up-front. The change is also likely to have some unspecified impact on revenue growth this year, but that should reverse significantly thereafter.
I think focusing on the growth of Confluent Cloud, and not focusing on getting large up-front commitments is the appropriate strategy. The growth of Confluent is all about getting users to adopt additional use cases, and it is not about getting users to make multi-year commitments. There is very little that sales people can do to change the curve of usage and much that can be done to identify new potential use cases.
This change will bring compensation into alignment with companies such as Datadog, MongoDB (MDB) and Snowflake (SNOW) and the hyperscalers. It makes sense for the company to institute the change in a year in which its likely growth was going to disappoint in any event.
The company is now going to shift forecasting its expectation for subscription revenues both quarterly and for the year as a whole, although I believe the company will still report RPO and cRPO balances for some time.
Most often when I write these deep dive articles I go through a detailed analysis of the business model. In this case, that really isn’t appropriate-I expect that this will be a very different company in 2025 with sales incentives retooled, and different cost ratios.
The company has been achieving quarterly records in non-GAAP gross margins; last quarter that metric was 76.4% which compares to 70.9% in the year earlier period, and to 74.7% in the prior sequential quarter. The company believes that non-GAAP gross margins will continue to rise in 2024 despite a mix swing to cloud which impacts that metric. Overall, the company, despite the reduced growth estimate is estimating that full year 2024 results will reach a break-even level or be marginally profitable both in operating margins and free cash flow-a 900 bps improvement overall.
Wrapping up – Reiterating the case to buy Confluent shares
Confluent reported reasonable results for its 3rd fiscal quarter but its guidance lagged prior expectations and the shares fell by ~40% in a single day. Since then the shares have strengthened in line with the overall market and the shares were up by 9.6% today, November 14, 2023.
The company’s weaker than anticipated forecast was a function of a slowdown in usage coupled with a slowdown in new use cases built on the Confluent Cloud. These slowdowns seem to be macro related, similar to what many other companies have seen in the recent past. There is no indication that potential users are looking at competitive alternatives although given Confluent Cloud is based on open source Kafka it is always possible that some users are choosing the free version despite the many problems with that for an enterprise deployment.
Confluent is the current leader in stream processing. Its founders developed Kafka, the de facto standard in stream processing while at LinkedIn. Subsequently two of the Kafka developers founded Confluent which offers an implementation of Kafka that has many enterprise grade features and is easier for most developers to work with.
Earlier this year, the company acquired Immerok. Immerok has developed a paid version of Flink, an open source data processing platform. The revenues of Immerok were inconsequential when it was acquired. The company has already announced the combination Confluent Cloud and Flink into a single combined solution. This will be in general availability before the end of 1H 2024 and should have a palpable impact on revenue growth in 2025.
The company is changing its sales compensation structure to align incentives with company business goals. Rather than incenting multi-year commitments, the company is going to incent its sales staff to maximize the growth in the company’s subscription revenue. While this may have some short term headwinds as the change is implemented, aligning goals of sales people and Confluent makes lots of sense.
The use of Kafka and Flink as a foundation technology for application development is likely to expand rapidly after the current hiccup. The ROI for users is huge and the ability of the technology to facilitate a positive user experience is substantial.
No one investing today in Confluent shares is doing so because of the 22% growth and marginal profitability the company is projecting for 2024. What they are investing in is a strong growth rebound in percentage terms in 2025 and beyond coupled with a positively evolving business model.
It is my contention that by the time the rebound is in clear view, the valuation will already more than reflect that expectation. I realize that by recommending the shares of Confluent at this time I have adopted a bit of a contrarian stance. I think that in a high growth IT portfolio, especially in this environment, a certain weighting should be reserved for companies whose prospects are bright, but whose current performance is limited by economic headwinds.
Realizing that patience is likely required, I believe that over the course of the next year, Confluent shares will produce positive alpha.
