Jeremy Burton is a seasoned Silicon Valley veteran, used to managing vast resources inside giant companies such as Dell, Oracle and EMC. He has seen many waves of change in the Valley.
What is going on with Snowflake, however, is perhaps on another level. “The main thing is how many new use cases there are, the ecosystem is starting to get going,” he reflected recently in an interview with ZDNet via Zoom.
Burton’s startup company, Observe, Inc., is just 60 employees, and is still refining its product. Already, though, Observe has a front-row seat to the Snowflake revolution as one of the big users of Snowflake.
Among the things that blew away Burton is that Snowflake has been able to build a data store that can perform traditional table joins across enormous data sources using plain-old SQL query language statements.
“The thing that blew me away a little bit about Snowflake’s success was that it was SQL,” he recalled.
“Even ten years ago years ago, it was like, yeah, are people really going to use SQL,” he observed. “Turns out that everybody still wants to use SQL.”
The magic of Snowflake, to Burton, is being able to create databases of unparalleled scale.
“If I distill it down, relational databases, they’re masterful at relating data, it’s in the name; what Snowflake does is, you can do these relationships of not just things that have a thousand rows or a hundred thousand rows, but a billion rows.”
The result, he said, is “Joins at scale, if you’ve got a million rows in one table, which is all your traces, and a million rows in another, which is all of your logs, we can do that join, and that’s just never been possible before.”
The broader information awareness, if you will, is key for Observe, which is building a tool for observability, which it intends to be sort-of “the one analytics to rule them all,” as ZDNet has written.
“These huge joins, for me, it’s about all about providing the user with context,” explained Burton. “If you’re looking at this, what else can you see that is related.” That relationship part “is critical,” he said. “We lean on the database to do that.”
Which table rows to join, of course, is also important, and that is the art of what Observe is building. “The Snowflake code that we generate is highly efficient,” he observed. “We know how Snowflake operates,” he said.
Observe is the tip of the spear, as it were, in the Snowflake advance, the kind of application that really pushes the system’s capabilities.
“We’re probably in the top 0.1% of folks who really push the platform, the kind of things we’re doing, not many people in their customer base are doing.”
That can lead to “tension points,” places where one company needs to press the other for what they want, he noted.
“Obviously, there are things in the platform that we would like to be done better,” he said, referring to Snowflake’s platform. “We want their compiler to go faster,” to make queries go faster. “We send Snowflake the most horrendous queries they’re ever seen in their life!”
Snowflake has mostly optimized not for compile time but for execution time, because just speeding up SQL queries is actually a big payoff for Snowflake’s customers.
“Most of their customers are these big banks that run these big jobs that used to take two days, now they take twenty minutes, if the compile time is a minute in twenty, they don’t care.”
The core functionality of the product that has been built out first is logging, on the one hand, and metrics on the other hand. “The last six months, we’ve put all the work into handling metrics and associated alerts, time series data like DataDog,” said Burton, and logging like Elastic.
Also: One analytics to rule them all: Observe promises to resolve the muddle of Splunk, Datadog, etc.
Observe acts a replacement for Splunk. “The logging side of the product is very, very good,” said Burton.
The “first big Splunk take-out,” he said, was at startup Cybereason of Boston. “They’re probably one of the bigger companies we sold to,” at $120 million in revenue.
“They had Splunk to do security investigation and the cost was killing them,” he recalled. “We were able to go replace Splunk, and it wasn’t that hard.” Burton said Observe tends to run into more customers that are using Elastic for logging, especially the open-source version, rather than the commercial Elastic distribution.
When ZDNet pointed out to Burton that Splunk claims claims it is a BMW compared to less feature-rich competitors who are a Hyundai, Burton said Observe competes not only on price.
“I would say that people want a Tesla, and it’s cheaper than a BMW and it’s better technology,” says Burton.
Logging and application metrics monitoring are first up, and there is a third function that needs to be folded in, and a big one, which is tracing, the heart of application performance monitoring. “That’s not for today,” says Burton. “Our game plan is to do a great job of both logs and metrics, and then also take over APM.”
“Do we have all the functionality? Right now, no. But with the architecture we think you can get there.”
To Burton, the process of building out functionality in a deliberate manner is better than trying later on to “stitch it all together,” as he puts it.
“Bigger vendors like Splunk, Datadog and New Relic do what big companies do, you try to stitch it together,” he explained. But what ends up happening is, “you end up starting again, you end up with a new product.”
Burton was taking stock of some milestones for the four-year-old company. Over the past several months the company has gone from focusing strictly on engineering that product to now building out the sales team.
The sales team is being expanded from 15 to 20 people. Burton expects the customer list will expand from about 30 currently to perhaps 50 by year’s end.
Like Snowflake, Observe is recognizing revenue not on a fixed periodic schedule but rather as a consumption-based approach, where customers sign up for some amount of contracted usage, they use that at whatever rate they end up using it, and Observe recognizes revenue only as fast as the pace of that usage.
As such, the most relevant metric of the business at the moment, said Burton, is the annual contract value for a customer. He expects ACV will be at about $2 million by the end of this year.
“It’s been interesting in getting customers that are not familiar with paying for software on a consumption basis to be comfortable with that model,” said Burton.
Companies have apprehension on day one, he said, as to what their bill is going to be, and fears of skyrocketing bills. That’s an issue that Snowflake has had to deal with and now Observe has to tackle it as well.
What has been employed to ease concerns, he said, was starting out with contracts for small amounts such as $5,000 for 90 days to let the customer get a feel for it.
“If we’ve met the objectives, then we’ve proven that we can deliver something of value to you, and we’ll actually then have some history and usage.
“So then, when you go forward to do, let’s say, a buy that’s going to last you a year, we can predict what your bill is going to be.”
The positive picture for customers in buying this way is that it’s so little up front, it’s like “try before you buy,” rather than committing to a traditional software term license from the outset.
“All of the usage data we’re going to put in the product,” said Burton, So That customers can know precisely what they are consuming, he said. “It’s the antithesis of enterprise software in days gone by,” said Burton.
Snowflake also gives usage data, for things such as storage consumption. But Burton aims for Observe to go a step further, giving details on the data sets that are being queried, the cost of querying that particular data set, and the individual user doing the queries.
As a result, he expects ACV sizes will steadily rise as utility is demonstrated. “The deals that were $4,000 to begin with turn into $30,000, $40,000, $50,000, and then to half million but only if we help the customer turn that into something successful.”
Burton, who had run applications at Oracle, sees in Snowflake’s expanding ecosystem an echo of what happened at Oracle.
“The first act at Oracle, we sold licenses, but the second act was partners, folks like SAP and Peoplesoft and Siebel.”
“Oracle printed money, because whenever SAP sold an application, it just drove a whole bunch of Oracle sales, and Oracle, they just sat back and picked up the money.”
Similarly, he argues, for Snowflake, “the next act for them is enabling folks like us.”
Burton looks forward to a day when Observe will be important enough to be minting money for Snowflake.
In addition to running a company to build on top of Snowflake, Burton is on the board of directors of Snowflake. That must make things easier for Observe, Inc., you would think.
Not necessarily so, he said.
“The talk in the board meeting is not always about the nuances of the compiler,” said Burton with a laugh.
“The hope is as we [Observe] get bigger and more relevant, they pay more attention to us.”
