By Peter Bailis - October 16, 2019
Organizations live and die by their performance on two metrics: profit and loss. However, these metrics are backwards-looking – they reflect a long and complex history of business decisions spanning multiple teams, from R&D to sales, marketing, and finance. It’s often hard to tell how an organization is performing until the window for making critical decisions has already passed.
As data has gotten easier to collect, we can afford – for the first time in history – to track almost every business metric at increasingly fine granularity. And we aren’t just recording more metrics – we’re also capturing more context for every event, like customer demographics, loyalty programs, and even weather conditions. As a result, a modestly-sized enterprise data warehouse today contains more data than the entire Google search index did at launch in 1997. With this data, every team can make data-informed decisions, and every business operator can become an analyst – in theory.
Despite these amazing advances in data availability, it’s still too hard for business operators to answer one of their most important questions: why is a given metric changing? Today’s core analytics toolkit – namely BI tools and SQL querying – lets users curate dashboards, build reports, and manually explore their data. These tools make it easy to see how a given metric is performing at one point in time, but it’s near impossible to tell quickly or with certainty why a metric is changing.
Imagine you’re responsible for the launch of your company’s biggest product yet. You’ve invested hundreds of millions of dollars in R&D, and now it’s up to you to coordinate between product, marketing, sales, and operations as you launch globally. This morning, you checked the launch dashboard and saw that new user activations slowed dramatically in the last 24 hours. Why?
There are literally millions of possibilities. Are activations down in US? Canada? Across the board? Is the dip due to users in store? Online? On mobile? Is the new marketing campaign underperforming? Among teens? Among the elderly? Is there a bug in the product? Or is it some combination of these factors? Depending on the answer, you’ll need to take a different set of actions and involve different teams.
Accurately diagnosing a change in a single metric can take days, and sometimes weeks. There are too many possible factors and combinations of variables for a single person (or team of analysts) to check, and the answers presented are often inaccurate, and biased. What’s more, as soon as the data changes, so do the answers. The few analyst teams that are dedicated to diagnosing these changes face a Sisyphean task, and are chronically understaffed. Even among the best-funded organizations, it’s impossible to scale people to keep up with the rate that the data is changing.
We started Sisu to bridge the gap between the availability of data and its use in everyday decision-making. Our goal is to enable everyone to understand why their key metrics are changing – in real time, and using all of their data.
Sisu monitors KPIs for changes and identifies the key drivers behind those changes, proactively notifying users when new facts arise. This allows our users to understand what’s affecting their businesses and to make decisions with confidence. It’s the fastest and most comprehensive diagnostic platform for structured data.
Today, it’s easy to take it for granted that anyone can search trillions of documents on the public Internet in seconds; in fact, if you can’t find something on Google, it probably doesn’t exist. We think it should be just as easy to leverage an organization’s private, structured data to continuously inform key decisions, and automatically diagnosing key metrics is a critical first step.
We spent years in my lab at Stanford building new interfaces, ranking algorithms, and scalable systems to analyze metrics derived from structured data. After developing and deploying prototypes with collaborators like Facebook, Microsoft, and Google, we realized both the fundamental need and the scarcity of tools available to diagnose metrics over extremely large and dynamic structured datasets. Most of today’s organizations already have enough data to diagnose change in their key metrics in real time; we believe every company that’s still around in 20 years will be doing so daily.
Today, we’re announcing the general availability of the Sisu platform. Over the past 14 months, we’ve assembled an amazing, multidisciplinary team that spans world-class user experience design, machine learning, and database engineering to build a product that couldn’t exist without all three. In stealth, we’ve iterated quickly with early customers and built a product experience that’s unlike any other. And market demand has exceeded our expectations – we’re thrilled to count organizations like Samsung, Upwork, and Mixt among our customers at launch.
Operational analytics is a wide open space, with a unique set of product and computational challenges that demand a fundamentally new approach. To capitalize on this market opportunity and maximize our impact, we’ve raised a $52.5M Series B led by Pete Sonsini at NEA, with participation from Green Bay Ventures, the a16z Cultural Leadership Fund, and our existing investors at Andreessen Horowitz. I’m thrilled to be working with both Pete and Ben Horowitz on our board – they combine decades of experience in enterprise software, SaaS, and data analytics. We’re already putting this capital to work scaling our team and the platform to make Sisu even faster and more useful, changing the way our customers use data to change the world.
Update: Read what Pete and Ben have to say about the future of analytics: