Somewhere, a dashboard is loading. Nobody is watching it.
Every company builds them. Slide decks full of charts, a BI portal with forty tabs, a quarterly report that renders a spinning wheel while an analyst apologizes. The dashboard is the most-purchased and least-opened software in the enterprise. It waits. It buffers. And the people it was built for quietly go back to asking a colleague over Slack.
Rill Data started with that unglamorous truth. Michael Driscoll, who had already spent a decade in databases - founding Metamarkets, watching it get acquired by Snap in 2017, and helping build the engine that became Apache Druid - looked at the state of business intelligence in 2020 and reached a blunt conclusion: it was broken. Not missing features. Broken in the way that matters, which is that people did not use it.
His fix was not another chart type. It was speed, and a change of philosophy. Rill packs a last-mile ETL service, an embedded in-memory database, and operational dashboards into a single tool. Point it at a data lake - or an ordinary S3 bucket - and it returns a live, sub-second dashboard. No cloud warehouse required. No ticket to the data team. The database disappears; the metrics stay.
The other half of the idea has a name that sounds like jargon until you watch it work: BI-as-code. Your dashboards, dimensions, and business logic are written as SQL and YAML, versioned in Git, reviewed in a pull request, and shipped like any other feature. Analytics stops being a portal you log into and becomes a codebase you build. And in 2024, that same code-first design turned out to be exactly what AI agents needed to read from too.
Above: the Rill mark, framed. A rill is a small, fast-flowing stream - which is either a coincidence or the most on-brand company name in data infrastructure.
Rill is the fastest business intelligence tool for humans and agents.
Four moving parts, one tool
Rill Developer
An Apache-licensed BI tool that runs on your laptop with an embedded DuckDB engine. Connect to a lake or warehouse, model in SQL, and get an exploratory metrics dashboard in seconds - no cloud account to start.
Rill Cloud
The deployment layer, powered by ClickHouse and Apache Druid. Serves fast operational dashboards to whole teams, with Git-backed deployments, CI/CD, and row-level access policies.
BI-as-Code
Dashboards, dimensions, and measures defined as SQL + YAML artifacts. Version them, review them, deploy them. Metrics-first, with SQL at the center instead of a drag-and-drop maze.
Conversational BI + MCP
Natural-language querying plus an MCP server that wires AI agents directly into live metrics - so humans and agents read from the same source of truth.
Connect
Point Rill at S3, a lake, a warehouse, or a stream. 20+ connectors.
Model
Shape data with SQL. Define metrics and dimensions as code.
Explore
Sub-second slicing, filtering, and drilling on an embedded OLAP engine.
Deploy
Push to Git. Ship the dashboard like software. Agents query it too.
The whole pitch in four steps: the fastest path from raw data to a metrics dashboard your team opens without being asked.
Built for a metered world
Programmatic advertising moves in milliseconds and bills by the impression. Fintech reconciles transactions that cannot wait for an overnight batch. IoT fleets emit telemetry by the firehose. These are the workloads where a dashboard that takes ten seconds to load is not slow - it is useless.
Rill processes more than 100 billion daily events across thousands of users, with named customers including Bloomberg, Comcast, and AT&T, alongside a growing bench of fintech and ecommerce teams. In programmatic advertising in particular, Rill has become a fixture among audio SSP clients - reportedly requiring little to no training, which for BI software is close to a miracle.
The common thread is real-time, operational analytics: instant drilling and dicing on live data, and increasingly, fast conversational agents that answer questions directly against it.
Raised on a one-line thesis: business intelligence is broken, and the fix is fewer moving parts, not more features.
Michael Driscoll keeps rebuilding the same layer
Metamarkets. Then Snap, after the 2017 acquisition. Then Rill. Along the way, the engine his team built was open-sourced as Apache Druid and spread across the industry. Driscoll did not wander into analytics - he has been circling the same problem for over a decade, each time getting closer to the part that actually matters: the last mile, where a query becomes an answer a human can act on.
When he started Rill in 2020, he brought engineers he had worked with at Snap. The team stayed small and deliberately technical - the kind of group that argues about OLAP engines for fun and then picks DuckDB because, on a single node under 100GB, it is routinely benchmarked as the fastest analytics database anywhere.
Business intelligence was broken - so we rebuilt it as code.
Michael Driscoll founds Rill with the premise that BI is broken, recruiting engineers from Snap.
Emerges with a $12M seed round to reimagine business dashboards.
Open-sources Rill Developer and deepens the DuckDB integration for local, exploratory analysis.
Publishes joint case study with ClickHouse on real-time operational BI for programmatic advertising and metered workloads.
Positions as agent-first BI: dashboards authored by AI agents in YAML/SQL, with an MCP server connecting agents to live metrics.
Notes from the margins
See it move
Somewhere, a dashboard just loaded. Someone is watching it.
Return to that opening scene - the buffering chart, the apologizing analyst, the colleague getting pinged on Slack instead. Rill Data's whole reason for existing is to delete it. When the dashboard answers in under a second, people stop working around it and start working with it. Speed, it turns out, is not a luxury feature. It is the difference between software that gets used and software that gets bought.
And the scene has changed shape again. The person watching the dashboard might not be a person. It might be an agent, querying the same semantic layer, reading the same code-defined metrics, acting on the same live data. Rill built for that on purpose - which is why a small team in San Francisco, processing a hundred billion events a day, keeps quietly rebuilding the least-glamorous, most-important screen in the enterprise.