Sisu secures $ 62 million to expand its data analytics business
Sisu Data, a spin-off of a Stanford DAWN project that puts AI to work in the “combinatorial explosion” of deciding which data variables to track, today announced it has raised $ 62 million in as part of a Series C investment cycle. The company is also rolling out a new dashboard and new data mining capabilities in its offering.
As explained by Peter Bailis, CEO and Founder of Sisu Data, Datanami last year, Sisu is based on a research program he led as an assistant professor at Stanford DAWN. (Bailis, who was the founder of the DAWN Project, left Stanford in 2020.)
“What we found was that in many cases the challenge wasn’t just figuring out what was going on – any BI tool or environment can do it,” Bailis explained to us. “When you have this super large data [with] all these different features and columns and so on, figuring out why the metrics are changing ”is the hard part.
If you knew exactly which SQL query to run (not to mention perfect data), then it would be easy to understand the root cause of a given metric rising or falling, according to Bailis. But without precognitive skills, which are rare in this world, analysts have to “brutally force” this type of analytical activity by generating many assumptions, turning them into questions, and refining their SQL queries until they get answer. .
This idea forms the basis of Sisu’s offering. According to Bailis, a Datanami 2021 Person to watch, Sisu essentially reverses the problem. Instead of waiting for a metric in a dashboard to change and then tasking an analyst to figure out why that is changing, Sisu uses machine learning to constantly monitor the entire data space and escalate automatically. the interesting elements.
“You declare the metric. You tell us the attributes. We’ll go do the slicing, ”Bailis said in the 2020 interview.“ It’s basically running some really big hypothesis tests, statistical tests, to figure out which variables are interesting and important, and then [what] are ways in which you can transform these variables.
Think of an OLAP cube with the ability to slice and slice data across multiple variables, but without needing to actually materialize the data (as that would be too slow). “Because we do a lot of these irregular groupings, we basically have a mostly memory-based columnar MPP parallel data flow engine, which basically does some smart data encoding and a bunch of… parallelization and so on, that which gets us going from a material point of view, so fast, ”Bailis told us last year.
Over the past year, Sisu has nearly doubled its workforce and more than tripled its turnover. It has clients with names like MasterCard, Autodesk, Samsung, Upwork, Wayfair, Equinox, Udacity, and Gusto.
It is now looking to step up its time to market with the $ 62 million Series C funding led by Green Bay Ventures and with participation from a16z, NEA and new investor Geodesic Ventures. The round brings the total risk-taking of the San Francisco, Calif.-Based company to $ 128.7 million.
“Sisu is creating a new way for organizations not only to analyze their data, but also to use it to make the best decisions to improve operations, profitability and the success of their business,” said Anthony Schiller, Co-Director General of Green Bay Ventures. A press release.
Sisu also today launched Explorations and Dashboards, two new features designed to help businesses better visualize and understand their data.
“Explorations allow users to quickly and easily dig, rotate, and view their metrics, without the need for code,” Bailis said in a blog post today. “Dashboards allow users to collect crawls, track changes to their metrics, and share them with others. “
Sisu noticed that users still used PivotTables to explore their data and then switch to Sisu to do more detailed analysis. That will change with dashboards, which put ML-based diagnostics at the user’s fingertips, “so you can automatically reveal the main drivers of changing metrics with one click,” writes Sisu’s Davide Russo in a separate blog post today.
Discover the crystal ball of advanced analytics
VC Ben Horowitz talks about Hadoop, AI and data culture
Sisu seeks to answer why
Editor’s Note: This story has been corrected. Peter Bailis was an assistant professor at Stanford, not an associate professor. Datanami regret the error.