A member of our Rising MAD Index of corporations on their path to an IPO, Confluent is a really attention-grabbing firm in a strategic a part of the info house, offering infrastructure for real-time information streaming – what it properly calls “information in movement”, in distinction to the world of batch processing or “information at relaxation”.
I had the pleasure of internet hosting the corporate’s co-founder after which CTO, Neha Narkhede, at Information Pushed NYC again in 2016, and her nice speak stays solely related to know the premise behind the corporate and its core technical basis.
Confluent lately launched its full S-1, and can commerce below the inventory ticker CFLT on the NASDAQ.
In the identical vein as earlier “Fast S-1 teardowns” (see Palantir, Snowflake, nCino), listed here are some excessive stage ideas and fast highlights, from my colleague John Wu and I.
HIGH LEVEL THOUGHTS
The streaming alternative and past:
- Actual-time information processing has been a scorching subject for the reason that early days of the Huge Information period, 10-15 years in the past – notably, processing pace was a key benefit that precipitated the success of Spark (a micro-batching framework) over Hadoop MapReduce
- Nevertheless, for years, real-time information streaming was all the time the market section that was “about to blow up” in a really main method, however by no means fairly did. Some trade observers argued that the variety of purposes for actual time information is, maybe counter-intuitively, pretty restricted, revolving round a finite variety of use circumstances, like on-line fraud detection, internet advertising, Netflix-style content material suggestions or cybersecurity.
- Confluent has actually proved the naysayers flawed, because it’s clearly constructed a powerful, fast-growing enterprise
- The bull case for Confluent in public markets is that the shift to real-time information processing is simply beginning in earnest:
- On-line machine studying, IoT and microservices are drastically growing the necessity for actual time, as the corporate notes
- The vary of use circumstances will proceed to increase, not only for actual time information processing, but in addition, more and more, actual time information analytics – powered by a rising stacks akin to Apache Kafka + ksqlDB + ClickHouse (or Druid, and so on) + Superset (or Looker, Mode, and so on.).
- Confluent’s ambitions transcend real-time information:
- Confluent desires to regularly conquer “information at relaxation” use circumstances as effectively. KsqlDB, its native occasion streaming database, “unifies the processing of knowledge in movement and information at relaxation”. Confluent believes it should result in vital displacement of batch information processing on conventional databases and a corresponding shift in spend to information in movement applied sciences.
- In consequence, Confluent views itself as changing into a foundational information platform within the enterprise for all use circumstances: “the central nervous system of a company, permitting information to be captured and processed as it’s generated round the entire group, enabling organizations to react intelligently in real-time.”
- Will it get there?
- There’s no lack of different corporations within the ecosystem that additionally view themselves as the center of the enterprise’s information infrastructure (e.g., Snowflake, Databricks). As well as, the streaming world itself is more and more aggressive with each giant gamers (Azure Occasion Hubs, AWS Kinesis and DynamoDB Streams, Google Dataflow, TIBCO Streaming, and so on) and rising startups (for instance, see our latest hearth chat with Materialize)
- That is actually an unlimited market, with room for a variety of profitable public corporations, as we’re within the early innings of the deployment of knowledge infrastructure as a core basis for any firm around the globe
- The important thing query whether or not Confluent’s spectacular efficiency within the information streaming world can change into a beachhead for a broader domination of that big market, the place it could exchange over time loads of different key repositories.
Open Supply and the Energy of Kafka
- Confluent is one other instance of an thrilling fashionable enterprise software program firm constructed on prime of open supply – Apache Kafka
- That is additionally an instance of the now traditional enterprise startup creation story that makes VCs swoon: sensible engineers (Jay Kreps, Neha Narkhede, Jun Rao) work at an enormous Web firm (LinkedIn) and are confronted with technical issues that the remainder of the world has not skilled but. To resolve these issues, they create a brand new framework (Kafka). They open supply stated framework (in 2011), which quickly positive aspects in reputation. Just a few years later (2014), they depart the large Web firm to create a startup (Confluent), which can construct the business model of the open supply software program. VC cash pours in, and the corporate is off to the races.
- It’s attention-grabbing to notice that Confluent to at the present time stays fairly single threaded on Kafka, significantly for those who evaluate it to an organization like Databricks, which rose to prominence on the again of Spark, however has since then launched a number of of open supply merchandise (MLFlow, Koalas, Delta Lake, and so on)
- Kafka is an enormous a part of Confluent’s bottoms up, developer first go to market movement
- Kafka can be an enormous a part of Confluent’s moat
- There are actual challengers to Kafka nowadays, particularly the mixture of Apache Pulsar and Apache Flink, which many argue affords larger efficiency.
- Nevertheless, Kafka is difficult to displace due to the scale of its developer neighborhood (“greater than 60,000 meet-up members throughout over 200 world meetup teams, estimated to have been utilized by over 70% of the Fortune 500“) and the general maturity of its ecosystem (companions, connectors to a whole bunch of sources, and so on.)
- Like different open supply corporations (Redis Labs, MongoDB, Cockroach Labs, Elastic), Confluent launched some restrictions on its open supply mannequin, again in 2018. This was to guard itself in opposition to the large cloud suppliers, akin to Amazon which had simply introduced its personal absolutely managed model of Kafka. Confluent didn’t make any modifications to the Kafka license itself. As an alternative, it created a brand new Confluent Group License that’s “considerably differentiated from Apache Kafka and was basically re-architected to function at cloud-scale, whereas being interoperable with present Apache Kafka techniques”.
- Confluent adopted the identical path to monetization as different profitable open supply corporations, going from an open supply mission to an enterprise self-managed providing (Confluent Platform) to a totally managed providing (Confluent Cloud). The corporate is clearly betting that Confluent Cloud will change into to Confluent what Atlas has change into to MongoDB (the place it went from a brand new product in late 2016 to representing 46% of MongoDB’s income in 2021, see our hearth chat with Dev Ittycheria, CEO of MongoDB). Whereas Confluent Cloud is the quickest rising a part of Confluent (see beneath), it’s nonetheless early days because it represents “solely” 18% of Confluent’s total revenues.
- FY20 ended with income at $236.6M
- FY20’s internet loss was excessive at ($229M)
- $280M of money on the steadiness sheet as of March 31, 2021 – not that a lot given the scale of the online loss
- 1,473 workers working throughout 20 nations
- Based in September 2014, the corporate grew extraordinarily shortly by way of most of its existence:
- Broke the $100M annual recurring income mark in just below 5 years in April 2019
- From 2018 to 2020, whole income grew from $65.2M to $236.6M
- Nevertheless, the tempo of development has slowed down considerably, albeit on a bigger base: +130% YoY from 2018 to 2019 however solely +58% YoY from 2019 to 2020
- This actually continues to make Confluent a excessive development firm, however far beneath some finest of sophistication growers at IPO like Datadog or Snowflake (which for instance was rising 174% yearly to $264.7M the total fiscal yr earlier than its IPO)
- Based on the administration dialogue, COVID-19 represented solely a modest opposed affect on sure elements of the enterprise
Strains of Enterprise:
- Two broad product choices: 1) Confluent Cloud – fully-managed cloud-native SaaS providing, and a pair of) Confluent Platform – enterprise-ready, self-managed software program providing that’s cloud-agnostic and multicloud, and might be deployed on premise, non-public cloud, or within the public cloud
- Subscription choices (Confluent Cloud, Confluent Platform licenses, and Confluent Platform publish contract help, upkeep, and upgrades) is rising sooner than providers, consisting of 88% of income up, from 87% in 2019, whereas providers made up 12% in 2020
- Confluent Platform development was largely pushed by PCS (publish contract help, upkeep, and upgrades) reasonably than licensing, rising 63% from $78.7M in 2019 to $128.2M in 2020
- Licenses for the Confluent platform grew simply 32% from $37.1M in 2019 to $49M in 2020
- Confluent Cloud, first launched in 2017, is the quickest rising Confluent product, and is prone to be the most important future development driver.
- It grew 117% YoY from 2019 to 2020, vs + 53% YOY for the Confluent Platform
- Nevertheless, it’s nonetheless nascent contributing solely 13% ($31.4M) of Confluent’s income in 2020, up from simply 10% in 2019. For the primary quarter of 2021, this determine was 18%
- The corporate talks loads within the S-1 about its capability to “land and increase” at clients and develop over time, however its Internet Income Retention (which measures upsells to clients) is lowering
- Greenback-based internet retention for 2020 was 125%, drastically lowering from their prior best-in-class price of 177% in 2018 and 134% in 2019.
- The decline appears primarily pushed by:
- the affect of present clients changing into a bigger portion of each the general buyer base and ARR,
- giant preliminary deal sizes that incorporate potential development,
- the affect of the COVID-19 pandemic, and
- the preliminary affect of present clients transitioning to Confluent Cloud
Gross and Internet Margins:
- Gross margin decreased from 75% in 2018 to 67% in 2019, earlier than enhancing barely to 68% in 2020
- This lower was pushed by lowering margin for throughout each subscription (83% in 2018 to 76% in 2020) and providers (down from 19% in 2018 to six% in 2020)
- Progress in working bills from 2019 to 2020 (+99% YoY, pushed by a spike usually & administrative bills) outpaced income development (+58% YoY)
- Confluent’s internet margin in 2020 was -97%, down from -64% in 2018 and -63% in 2019. This lower might be attributed to growing OpEx throughout all capabilities over the two yr interval, with ballooning R&D (+157% YoY) and S&M (+112% YoY) spend in 2019, and G&A coming in 2020 (+397% YoY)
- As of March 2021, Confluent had 2,540 clients (up from 820 in December 2019), serving 136 out of the Fortune 500
- Confluent has 513 paying over $100K in annual recurring income, and 60 paying over $1M ARR
- Notable clients span throughout industries, together with Goldman Sachs and Morgan Stanley in monetary providers, NASA JPL and the Facilities for Medicare & Medicaid Companies in authorities, Michelin and SunPower in manufacturing, Advance Auto Elements, Dick’s Sporting Items, and Domino’s Pizza in retail, and Instacart, ServiceNow, and Seize in know-how
- Whereas Confluent’s absolute development is essentially pushed by growing US gross sales (+$54M from 2019 to 2020), Confluent’s worldwide presence is increasing at a sooner price.
- In 2019, 32% of Confluent’s income was from outdoors of the US. For 2020, that determine was 34%, and for first quarter of 2021, it was 36%
- Confluent’s worldwide income grew 67% YoY from 2019 to 2020, in comparison with simply 53% for it’s American section
- No nation represented greater than 10% of Confluent’s income except for the US throughout any interval
Funding & Possession
- Based on CB Insights, Confluent has raised $455.9M throughout 5 rounds, largely lately elevating a $250M Collection E led by Coatue, valuing the corporate at $4.5B. Different buyers embody Sequoia Capital, Index Ventures, Benchmark, Franklin Templeton, Altimeter Capital, Information Collective, and LinkedIn.
- Founder possession: Jay Kreps owns 12.6% of excellent shares, Jun Rao owns 10.6%, and Neha Narkhede owns 1.6%.
- Investor possession: Benchmark owns the biggest stake at 15.3%, adopted by Index with 13%, and Sequoia Capital with 9.3%.
DISCLAIMER: none of that is funding recommendation — for enterprise capitalists like us, going by way of S-1s is a routine studying train, and we’re simply “open sourcing” our effort in case it may be attention-grabbing to others.