A little over a year ago, we purchased the Metamarkets business from Snap, with the intention of bringing this amazing data visualization product to markets beyond advertising.
While running MoPub, Metamarkets was a key tool that unlocked a tremendous amount of value for us in our exchange. The founder, David Soloff, walked into our office in 2010 to demo a relatively complete visualization front-end for a typical ad exchange. At the time, we were handling maybe ten million ad requests per day … tiny! It was easy for them to handle our volume and show us our revenue, key metrics and more, in a straightforward exploratory user interface.
Just a couple of years later, our exchange had grown considerably. We were handling over a billion ad requests every day, each with dozens of high cardinality dimensions like country, device type, and publisher among others. Each ad request had several bid responses, each of which needed to be organized and aggregated.
In order to scale up to handle the load of MoPub’s exchange traffic, Metamarkets was forced to innovate. They invented a column-oriented data store called Druid, later open sourced, and now the basis for the Apache Druid project. Druid also inspired the Pinot project, created by LinkedIn and Uber, which Facet now uses for some of our modern analytics needs.
After I left Twitter in 2015, I joined the board of Metamarkets and helped Mike Driscoll sell the company to Snap in 2017, where it became a key driver of Snap’s revitalized ads business. We never forgot about the power of Metamarkets though. A few years later, Snap was wondering what to do with the business and we approached them with the idea of spinning it out.
I suspected that it would be a lot of work to reconstitute, given that most of the engineering team had moved on. But the core of what made Metamarkets such a powerful product remained: a flexible, fast, exploratory analytics product focused on revenue and operational analytics. It was a tool that business teams could use to make better decisions by observing streaming transactional data directly, rather than waiting for aggregated database tables to load.
We were able to negotiate and finalize a deal with Snap to purchase the Metamarkets business from them, and immediately got to work. We recruited a new management team and raised some additional capital from Harrison Metal, Bedrock and other friends in the industry, alongside the initial capital that Breakpoint Capital, my investing platform, provided.
The task in front of us was large. We needed to completely rebuild the product’s underlying data processing software using modern, cloud-based tools, with support for today’s popular data warehouse platforms: Snowflake and BigQuery. Yet we could not lose the magic that Metamarkets offered: the beautiful user interface front end. We modernized and rewrote the interface with a new design and modern tooling. Just a few months into the project, COVID hit us hard. We needed to realign from being an opinionated “office first” company based in New York City to a distributed company without a clear HQ almost overnight. We had to establish new asynchronous working habits, find a way to achieve team bonding virtually all while building new WFH setups and resisting the constant distractions of the stock market. (I will write more about these lessons on the Breakpoint blog soon.)
Only a little delayed, we finally are now taking the wraps off of the new Facet product, a beautiful, fast and capable data analytics product for streaming data, inspired by Metamarkets, a product that we loved and depended on all of those years ago. We’ve been working with a small handful of prototype customers over the last few months using Snowflake and we’re now ready to welcome more into the fold.