September 15, 2020
Snowflake Was Destined to Fail. What Happened?
Snowflake seems like such an obvious winner today with the likes of risk-averse Berkshire Hathaway taking an unprecedented stake in their IPO. But it wasn’t always the case.
I remember talking to several venture capitalists and industry insiders in 2013 and this was the consensus:
Snowflake is not a viable long-term company.
At best, people believed they could be a me-too provider of analytics in a world dominated by cloud leaders like AWS, Microsoft, and Google. They weren’t entirely wrong.
#1: AWS will kill them
The fact that Snowflake was built to run exclusively on the AWS cloud platform was considered the biggest risk. Just like Veeva on Salesforce, the conventional wisdom was that the platform owners like AWS (or Salesforce) could squish them at will.
It turns out platforms like AWS do want companies on their platforms to succeed--they win when more startups build on their platform.
#2: Hadoop will kill them
Snowflake was a closed proprietary system whereas the industry was headed towards open source and Hadoop was the clear winner. Here’s a news headline from 2013: With $141M in Capital, Cloudera sprints to $1B in Revenue.
It turns out that very few customers really care about their ability to touch and feel your source code. Even the best open-source companies are now pivoting to a cloud-first approach.
#3: Gross Margins will kill them
Snowflake pays a lot of money to AWS and the thinking was they can never walk out of that abyss.
It turns out that scale and pricing power can lead to great gross margins (61%). Here’s what Snowflake says about it in its IPO prospectus:
Our improved gross margin during the last four quarters presented is primarily attributable to higher volume-based discounts for purchases of third-party cloud infrastructure, increased scale across our cloud infrastructure regions, and improved platform pricing discipline.
#4: Low Switching Costs will kill them
Snowflake runs on data stored in AWS (or Google’s GCP or Microsoft Azure). This means customers can easily switch analytics providers to others that run on the same platform especially the platform owners themselves like AWS.
It turns out that while it’s true in theory--enterprises don’t really switch if they are being well served. In fact, Snowflake has a net negative churn; people buy more of it every year.
#5: Lack of Subscriptions will kill them
Unlike the SaaS players Salesforce, Workday, etc., Snowflake was not a subscription sale. You paid every month based on actual usage and not some pre-determined or pre-negotiated dollar amount. This means you can’t model 3 to 5-year Total Contract Value of millions like Veeva or Workday could.
It turns out that pay as you go pricing is an even more powerful model. As customers’ data increases and they run more queries, they end up paying a lot of money. Snowflake has some customers paying them north of $10M a year!
#6: Public cloud strategy will kill them
While people could envision storing sales leads or marketing emails in the cloud, it was conventional wisdom that the cloud was still not ready for prime time when it comes to financial services and healthcare companies. Why would a bank like Capital One put their most important data in the public cloud?
Snowflake runs on public clouds (only).
It turns out Capital One and many like them decided it was just too damn hard to compete against the tsunami of startups without being able to use the public cloud. They went all-in and never looked back.
#7: SaaS Analytics will kill them
This was my favorite. With all the cloud data sitting in the likes of Salesforce and Workday which offered their own analytics, why would companies bother to take a DIY approach? Wouldn’t all customer data end up in Salesforce, marketing data in Adobe, and employee data in Workday?
It turns out that few enterprises rely on just one SaaS provider. To get a sense of your business and your data, you have to combine data from internal and external systems, on-premise, and cloud apps, and so you do need a data warehouse.
#8: Sales & Marketing Costs will kill them
Snowflake is in a very competitive business. Likes of Google, Salesforce, and AWS are willing to spend a lot of money to get you to use their data platforms. This criticism rings true even today as they have to spend about a dollar in sales and marketing every quarter for every dollar in revenue they have. It literally burned through a $100M to get to its first few millions in revenue and has gone on to raise $1.4B so far.
It turns out that the market is willing to be very patient with big winners.
#9: Data Platforms will kill them
Snowflake is just a data warehouse solution whereas the companies it competes with like AWS, Azure, GCP all have end to end platforms that include not just the data warehouse but also relational databases, analytics tools, and much much more.
It turns out best of breed wins if you are truly better than your competition.
#10: Hired CEOs will kill them
Snowflake is an oddity in the startup world. We love companies started and run by founders. I am personally in that camp too. And here we have a startup that went out and hired a Microsoft executive as the first CEO. Heresy!
And then it fired Bob Muglia and hired another retired executive and seasoned CEO. Heresy!
It turns out that success comes in many shades, mostly blue. Snowflake blue is one of them. Salesforce blue is another.
I am hoping Skyflow blue is another one. We are learning form Salesforce (be cloud) and Snowflake (run on cloud). Skyflow is a simple turnkey solution to all your data privacy and residency needs--and soon you will be doing it right inside your Snowflake data warehouse.
With companies like Skyflow betting on the Snowflake platform, the future looks very blue.