Snowflake, before it was obvious

Peter Wagner
No items found.
At the time, there were several not-so-obvious aspects of our initial investment in Snowflake. They highlight the scale of the challenges the Snowflake team has overcome. They also illustrate some of the persistent biases that have a way of creeping into investment decision-making to this day.

Snowflake is now a public company -- an overnight success, eight years in the making! I first met founders Benoit Dageville and Thierry Cruanes in January 2013 (Marcin Zukowski was still in Europe) and led Wing’s investment in the Snowflake seed financing soon after. We have invested in every Snowflake financing since, right up to today’s historic IPO. But when we made our initial decision, things weren’t quite so obvious.

Before It Was Obvious

At the time, there were several not-so-obvious aspects of our initial investment in Snowflake. They highlight the scale of the challenges the Snowflake team has overcome. They also illustrate some of the persistent biases that have a way of creeping into investment decision-making to this day.

1.      Relational is dead

When Snowflake was getting going, fashionable thinking held that relational database technology was on its way out. Hadoop was all the rage for analytics. Intel would soon make an investment in Cloudera valued at $740 million. NoSQL technologies like Mongo were gaining rapid adoption for operational workloads, especially within the trend-setting startup developer community.

It was true that relational systems like Oracle and Teradata were struggling to meet the needs of many customers, especially rapidly growing “cloud-first” businesses. The founders of Snowflake felt that the problems lay not with relational technology itself, but with the way it had been implemented. They were certainly in excellent position to know, having been key architects of several generations of database products at Oracle and elsewhere. They believed that relational could be unleashed by rearchitecting it to take advantage of the power of the cloud. If they were right, the benefits would be enormous: the mainstream enterprise was already built around relational, with legions of analysts and DBA’s were fully trained in SQL, providing a large and ready market. But this notion was way out of step with the Silicon Valley belief system of the time.

2.      Amazon will kill you

In November 2012, Amazon released its Redshift cloud-based data warehouse into preview beta. This product would compete directly with Snowflake. Not only that, Snowflake was building its product on top of AWS computing and storage services. Amazon would be Snowflake’s chief competitor and primary supplier.

The founders of Snowflake understood the gravity of competing with Amazon, but believed they had key architectural advantages that would allow them to win. Redshift was built on technology licensed from ParAccel, a struggling data warehousing software startup. Snowflake was taking a completely fresh approach designed explicitly for the cloud and operated exclusively as a service. The Snowflake team was convinced that they could build a compelling lead in performance, functionality and economics over Redshift, which would be hamstrung by its legacy origins.

This product-centric argument met with a fair amount of skepticism at the time. Amazon had been mopping the floor with their competitors and had already dimmed the prospects of every major computing, storage and networking vendor. To most, it seemed inevitable that AWS would do the same with their growing lineup of data services. Even if the Snowflake product did turn out to be better than Redshift, few believed this would be enough to overcome Amazon’s ascendant market power.

3.      Enterprises won’t trust the Cloud with their data

This statement sounds crazy now, but it was once the conventional wisdom in the enterprise. In 2013, the enterprise data warehousing market was largely captured in on-premise deployments of Oracle, Teradata, Microsoft and a few others. Even the disruptive “Big Data” platforms like Cloudera and HortonWorks were deployed almost entirely on the customer’s own infrastructure (and still are, by the way). I remember attending a “Technology Summit” hosted by a top 5 financial institution in 2011 and asking a panel of their IT executives when they would adopt cloud-based analytics. They answered in unison: “Uh, never.”

The Snowflake team understood this hesitation but felt there was enough enterprise data “born in the cloud” to constitute a viable early market, and that the volume and importance of cloud-born data would ultimately dominate. They also knew that data security and data privacy would always be a paramount concern for their customers. Most observers felt these would be showstopper issues; Snowflake saw them as opportunities for differentiation and monetization.

4.      Who are these guys?

Benoit, Thierry and Marcin are exceptional, brilliant and determined technologists and entrepreneurs. They are among the most gifted individuals I have ever been lucky enough to work with. But in 2013, they were not exactly straight out central casting for the role of “Messianic Founder” in this Silicon Valley drama.

Benoit and Thierry had worked for many years deep inside Oracle, part of a tightly-guarded cadre of elite architects, the crown jewels of the company and a major reason it had held onto its market position for so many years. Marcin was still in Europe where he had worked nearly his entire life. They were not cool kids from a web-scale giant, nor were they open source legends. They were charming middle-aged engineers who just happened to understand the opportunity better than anyone else on the planet. But they didn’t fit the digerati’s mental model of where the next big thing would come from.

No alt text provided for this image

How It All Played Out

We already know this story has a happy ending (although Frank Slootman will be the first to remind you that we are at a waypoint, not an endpoint, and the journey is far from over). The team was clearly more right than wrong when it comes to the contrarian elements of the “Theory of Snowflake”.

1.      Relational technology has defied predictions of its demise, scaling to new heights thanks to the power of the cloud and products like Snowflake that have creatively harnessed it. Non-relational architectures clearly have their role as well, but the original “good ideas” of computer science that underpin relational, plus the enormous investments in tools and training already in place, still have tremendous sway in determining how enterprises derive value from their data.

2.      Snowflake has been able to compete effectively versus Amazon. Turns out that product excellence matters after all. AWS is a fierce competitor and understands the importance of the data layer, but there is a strategic rationale for customers to embrace a focused, cloud-neutral guardian of the “Data Cloud” that allows them to unlock the power of all their data assets and put them to use however they choose.

3.      Enterprises have embraced the cloud with a vengeance. Cloud-based analytics are becoming the dominant model, and Snowflake has emerged as a full-fledged Cloud Data Platform, supporting a range of workloads that goes well beyond its data warehousing entry point.

4.      Benoit, Thierry and Marcin, despite their “non-obvious” backgrounds, turned out to be exactly the right people to found this company. They also embraced help in the areas they needed it most from people like Mike Speiser (lead investor and original CEO), Bob Muglia, Frank Slootman and a host of other amazing team members that have made Snowflake a rare talent magnets.

The Upside Surprise

The four controversies discussed thus far turned out in ways largely favorable for Snowflake, which allowed the company to become successful. But what has allowed the company to become a true supernova was not as prominent in the early discussions. I’m talking about the Snowflake business model.

Those of us in the early Snowflake supporter camp always believed that the company could carve out a big chunk of the enterprise data warehousing market. And we thought that the best way to penetrate would be to make it easy for customers to try the Snowflake service. This is why so much of the development effort was devoted to making Snowflake super-easy to adopt and effortless to use – in stark contrast to the brain-exploding experience customers had come to expect from legacy data warehousing projects.

But even devout Snowflake fans like us were surprised by the alacrity of expansion. “Snowflake is addictive,” I was told by a Snowflake Sales Ops leader. “Once customers try it, they inevitably want more.” There was more pent-up demand for analytics in the enterprise than most of us realized, and Snowflake was breaking open the dam with its new model of delivery and consumption.

The company saw this happening and invested heavily in customer acquisition to fuel its “land” sales motion, a decision which ironically landed it in a tough spot. Metrics-oriented investors balked at Snowflake’s sales spending relative to new ARR acquired, and the 2016 financing had to be backstopped by the insiders. That was a mere four years ago, folks. The investors you see today tweeting Snowflake S-1 teardowns and buzzy factoids are not shareholders in the company and actually passed on that 2016 financing.

Today one of the most celebrated dimensions of the Snowflake business model is its superlative NRR (Net Revenue Retention) on quite substantial six-figure-plus ACV’s. As is the case with great consumption-based technology businesses, this property is baked into the product itself. It is the essence of Snowflake, and likely a hallmark of other great cloud-data businesses. I spend a lot of time with new startups in this ecosystem, and the potential for a supernatural Snowflake-like NRR is one of the key things I look for.

The Moral(s) of the Story

I’ve been doing this long enough to know that it is unwise to become too breathless. But, is there a bigger happening in enterprise technology than Snowflake? The company has become the foundation of the Modern Enterprise. At Wing, we define the Modern Enterprise to be an agile workplace built on data and powered by AI. The Cloud Data Platform (Snowflake being the archetype) makes it all possible.

A couple takeaways, informed by the experience of our journey with Snowflake:

  • Companies and investors can both be categorized as “Before It's Obvious” and “After It's Obvious”. Snowflake today is clearly “After Its Obvious” and has attracted a swarm of “After Its Obvious” acolytes. However, if it weren’t for the support of a core of “Before Its Obvious” stalwarts, we wouldn’t even be having this conversation.
  • Strategy is essential, but human factors carry the day. Benoit, Thierry and Marcin were not “fashionable” founders. But they were perfect for Snowflake, and part of their perfection was their eagerness to embrace contributions from others. They welcomed Mike as their original CEO, Bob who helped lay the foundation, and eventually Frank who is leading the way to greatness – in addition to the countless other Snowflakes who today are the company’s true competitive advantage.

It has been such an honor to work with all of you at Snowflake, and a joy to see your collective creation take shape over the past seven-plus years. All of us at Wing are thrilled by what you have accomplished and can’t wait to see what you will do next!

No alt text provided for this image

Wing Logo
Thanks for signing up!
Form error, try again.