Sigma

It is a Tuesday morning at a mid-size retailer, and a category manager who has never written a line of SQL is staring at every transaction the company made last quarter. Trillions of rows, technically. She sorts a column. She adds a formula. She types a question in plain English and an AI agent drafts the workbook for her. No ticket to the data team. No three-day wait. This is the ordinary miracle Sigma sells, and the strange part is how ordinary it looks.

The screen in front of her could be mistaken for a spreadsheet your accountant used in 1998. That is entirely the point. Sigma's whole argument is that the most radical interface for big data was the one everybody already knew.

Data went to the cloud. The people stayed behind.

By the mid-2010s, companies had solved the hard part of data: storage and compute. Snowflake, Databricks and the cloud warehouses could hold and crunch almost anything. The unsolved part was human. The data lived in a place only a handful of specialists could reach, fluent in SQL and patient with dashboards that took weeks to ship. Everyone else - the marketer, the finance analyst, the ops lead - filed a request and waited.

That bottleneck was good business for whoever owned the query language and bad business for everyone trying to make a decision before lunch. The dashboards were beautiful and the dashboards were stale. Asking a follow-up question meant starting the whole queue again.

Two builders, one stubborn idea

Sigma was founded in 2014 - originally incorporated under the unglamorous name Bitmoon Computing - by Rob Woollen and Jason Frantz, who started the company as entrepreneurs in residence at Sutter Hill Ventures. Woollen had come out of Salesforce. Their bet was almost contrarian: don't teach business people to think like engineers. Give them an interface they already understood and wire it directly into the warehouse underneath.

It took years to make a spreadsheet that could sit on top of a cloud warehouse and not buckle. The first customer signed in 2017. Sutter Hill backed an $8M Series A, then a $20M Series B in 2018 alongside Altimeter, and the headquarters moved into San Francisco's financial district. The unglamorous middle years of a company that would later look inevitable.

In 2020, in a move that tells you something about the founder, Woollen handed the CEO seat to Mike Palmer and stepped into the CTO role - so he could go back to building the product. Most founders cling to the title. He traded it for a keyboard.

A spreadsheet, but it never stores your data

Here is the trick under the hood. Sigma does not copy your data into its own system. It queries the warehouse directly, live, so Snowflake or Databricks or BigQuery stays the single source of truth. You get the comfort of a spreadsheet and the scale of a warehouse, without the usual cost of choosing one or the other.

On top of that foundation sits a growing toolkit: Workbooks that blend spreadsheet formulas, SQL, Python and charts in one document; Input Tables that let you write data back into the warehouse for planning and forecasting; Data Apps for building interactive applications without a separate stack; and embedded analytics so companies can drop Sigma inside their own products.

Then, in 2026, came the swing. Sigma Agents - no-code AI agents that run inside third-party warehouses, within the customer's existing security and governance. They answer questions, build workbooks, and now take action: interactively (you approve each step), autonomously (on a schedule), or externally (calling other systems by API). Sigma calls the category "agentic analytics." Two years ago, the category barely existed.

The numbers that doubled

Skeptics are allowed. So here is the receipt. In April 2026 Sigma reported annual recurring revenue had reached $200 million - roughly double the $100 million of a year earlier. A month later it raised an $80 million Series E led by Princeville Capital at a $3 billion valuation, itself double the $1.5 billion from the prior round. Strategic checks came from Databricks Ventures, ServiceNow Ventures and Workday Ventures, with longtime backers Altimeter, Avenir, D1, Spark and Sutter Hill returning.

The customer roster reads like a cross-section of the economy that has to make decisions fast: JPMorgan Chase, Duolingo and AMD among more than 2,000 companies. Over the prior year, Sigma added roughly 1.1 million more active users. The partnerships are telling too - Snowflake and Databricks are not just integrations but, increasingly, investors. When your warehouse vendor wants equity in the layer on top, the layer is doing something right.

Democratization, minus the buzzword

Strip away the slide-deck language and Sigma's mission is plain: let the person with the question reach the data without an interpreter. Governed, secured, auditable - but direct. The grandeur of "data democratization" lands better as something smaller and truer: fewer tickets, fewer waits, fewer people locked out of their own numbers.

It is worth noticing what Sigma did not do. It did not invent a new query language. It did not ask the world to learn its way of thinking. It took a 40-year-old interface almost everyone already knew and made it powerful enough to matter. The boldest move in enterprise software, it turns out, was to not make people change.

From answers to actions

The agentic bet is where this gets interesting. For thirty years, business intelligence stopped at the insight - here is the chart, now go do something. Sigma's agents are designed to close that gap, monitoring data and executing workflows, calling out to other systems, all inside the warehouse's guardrails. If it works at scale, the spreadsheet stops being a place you read your numbers and becomes a place where the work actually happens.

Back to that Tuesday morning. The category manager closes her workbook - the one no engineer touched - and moves on with her day. A decade ago, that question would have become a ticket, the ticket a queue, the queue a delay, and the delay a decision made on a hunch instead of the data. Sigma's contribution is not a dashboard. It is the missing minutes given back to the person who needed them.

Whether agentic analytics becomes a category or a footnote, the smaller thing already happened: the data warehouse learned to speak spreadsheet. And once a few million people get that, they tend not to give it back.