Made 200 million US dollars from BI

There is a pretty interesting thing. A company, with two successive leaders, has repeated the same sentence for eight years: We don't want to do BI.

But all the $200 million it earns comes from BI.

This company is called Sigma Computing, which develops analytics and BI tools running on cloud data warehouses; data does not need to be extracted, and queries are directly pushed back to the cloud data warehouse for execution, with main connections to Snowflake and Databricks.

Founded in 2014, it now has more than 2,000 customers and an annual recurring revenue of 200 million US dollars. Gartner has included it in the BI Magic Quadrant for two consecutive years, with a 4.8 score on user review platforms and more than 200 reviews.

For such a company, its founder and new CEO have said the exact same sentence for eight years.

In 2018, the company made its official debut. Founder Rob Woollen said, "Dashboards are not equal to intelligence." Note his wording: not equal to. Dashboards are useful, but they have reached their ceiling, and what he wants to build is far larger than dashboards.

In 2020, Mike Palmer took over as CEO. The first sentence he said after taking office was The path remains unchanged, and we will focus on scaling up.

Let's talk about his background. He previously served as Chief Product Officer at cloud backup vendor Druva, and before that, he was also the top leader in the product department at Veritas. He has a resume that is entirely rooted in data infrastructure from start to finish.

When a person with such a resume takes over a company, he can change the route at any time, but he did not, and said that he would continue to move forward along the original path.

In 2025, he appeared on MAD, a long-established podcast in the US tech circle. The host asked him what Sigma is actually doing, and he said two sentences:

The first sentence, the world does not need another BI tool. The second sentence is even more straightforward: they once shut down the original product, rebuilt it from scratch. It is not an iteration, nor an upgrade, but a complete overhaul.

In April of the same year, the company published a blog post celebrating that its annual revenue exceeded 100 million US dollars, which included two sentences: "We are not just another BI platform" and "The future of analytics will not be measured by dashboards".

In May 2026, the Series E financing was officially announced; Palmer's statement took on another form: customers voted with real money for Sigma, a place where they can build AI on top of their cloud data.

There is no mention of BI in this sentence. The official website's slogan, when translated into Chinese, is "AI Application Platform". It does not even want to be associated with the term BI.

Oh, by the way, there is another thing that is easy to overlook: after Woollen handed over the position of CEO, he did not leave the company. He has stayed until now, in charge of the product department. The founder stayed in the company and kept repeating this sentence, and the professional manager in the public eye also kept saying the same thing.

After eight years of development, the framework is being upgraded, from dashboards to BI and then to AI. The connotation is being narrowed down: dashboards are not enough, BI is not the end point, and in the end, it does not even want to be named after BI.

Well, two successive leaders, one sentence for eight years. Aren't you curious about what they are actually doing?

......

This question has to start from the time when they started their business.

The company was originally called Bitmoon Computing, with two founders, Rob Woollen and Jason Frantz.

At the time of starting the business, both of them worked at Sutter Hill, a US venture capital firm. While helping to review projects, they built their own prototypes and looked for directions. What were they looking for? An idea that allows people who cannot write code to query data to directly use the data stored in the cloud.

The idea is particularly correct, but they just could not find the right product form. In these four years, the team built prototypes and tried different directions. Some were discarded, and some were retained. Investors were sometimes unhappy, but they waited. The office was very shabby, but the advantage was that the cash burn rate was low.

It was not until November 2018 that it came out of stealth mode with its first product. The product form that everyone is very familiar with: spreadsheets.

Users operate data just like using Excel, formulas can still be used, and pivot tables can still be used. The difference is that the data does not enter the spreadsheet, but always stays in the cloud data warehouse.

As soon as you press the Enter key, the query runs on the data warehouse side, and the result returns to the cell. You do not need to write code, and you do not need to touch those other BI dialects.

There is also an origin story here:

Sutter Hill incubated another company at the same time, which later became the cloud data warehouse giant Snowflake. Mike Speiser, the managing director there, held two positions: he served as a director at Sigma and also served as the founding CEO of Snowflake.

Sigma later became so deeply tied to Snowflake, which had nothing to do with strategic choices, and the connection formed naturally over time.

Is this 2018 spreadsheet the thing they were looking for? Not yet. After that complete overhaul, the rebuilt product is much larger than a spreadsheet.

Where is it larger? To pick the most powerful feature: write-back. To put it in layman's terms, it covers both the hand that views data and the hand that modifies data; traditional BI, no matter how fancy it is advertised, only has one hand, which is read-only.

For example:

Traditional BI is like a car dashboard, which clearly displays speed, fuel consumption, and water temperature. But you cannot touch any mechanical parts through the dashboard. If you want to shift gears, you have to use another set of tools. Sigma's write-back feature opens up the other hand for you.

The feature is called Input Tables, which was launched in 2023. A table is placed in the dashboard, you can modify a number on the spot, and this data is directly written back to the data warehouse.

Business personnel fill in a column of forecasts, and all the charts change accordingly, without having to switch back and forth between a second system.

Is it that necessary? Isn't it just adding an input box in the spreadsheet? Of course it is. The trouble is that once the data is modified, the calibers, permissions, and audits on the data warehouse side all have to be adjusted accordingly. Among mainstream BI products, very few players dare to touch this feature.

Many people may ask, is this the same as the AI multidimensional tables that various companies are developing on the market?

Not quite. Most AI multidimensional tables modify the data in the cells of the table. Sigma modifies the underlying tables in the enterprise database, with calibers, permissions, and audits all adjusted accordingly. One modifies the surface, and the other modifies the source.

Competitors do not have this feature. If Microsoft's Power BI wants to write back data, it has to take a detour and combine two of its own other tools to make it work. Those established BI players like Tableau do not have native write-back at all. Among mainstream BI tools, Sigma is the only one that natively supports this feature.

Palmer later talked about the design logic in an interview: historical data is not the end point. Let business personnel fill in forecasts directly on top of historical data, so that the numbers are dynamic.

After viewing the data and modifying it, you can even directly trigger actions, which is exactly what Sigma wants to build.

There is also an English slogan on the official website. The first time I saw it, I thought it was a misplacement from a SaaS governance platform.

It says: Say goodbye to SaaS chaos and spreadsheet mess. It means that you don't have to make do with those small operation tools you subscribe to and those scattered spreadsheets everywhere, you can directly build formal data applications on Sigma.

The things built are called data applications in their own terms.

Looking at the product list in 2026: embedded analytics is available, AI Q&A is available, custom agents are available, and a notebook area that can run code is also available.

On the data source side, Databricks is added, and the data warehouses of Amazon and Google are also added. In August this year, it also signed a strategic cooperation agreement with AWS, and Databricks awarded it the 2026 Annual Partner Award.

The development pace of the AI line has also been moving forward.

In May 2024, it launched the AI toolbox. In August this year, it opened the interface for agents to fetch data, and in September, it upgraded the custom agent. Along the way, its position has never changed: it empowers people who view data, instead of replacing them.

Well, here is a harsh fact to mention: they want Sigma itself to become a module of the database, called Sigma Tables, which is still in beta since its official announcement.

That's all for the product introduction. The path is very clear: every new business added is attached to the main body of the dashboard. The main body is still growing, with more and more components added.

......

Can these new businesses be reflected in the accounts?

In May this year, when the Series E was officially announced, the company made all its cards public. Annual recurring revenue reached 200 million US dollars, the year-on-year growth rate more than doubled, there are more than 2,000 paying companies, and the number of new active users added this year exceeds 1.1 million.

200 million US dollars is equivalent to more than 1 billion RMB a year. What are these 1.1 million new people doing? I estimate that more than 90% of them open dashboards, not the input boxes for talking to AI.

Some people may ask, what percentage of the 200 million revenue comes from data applications?

To be honest, it cannot be separated. The official has never disclosed revenue by product line, and the pricing page does not even leave a way to split it. So who will fill this gap? I might as well calculate it with three points.

First, how the money is collected.

When customers subscribe to Sigma, the main thing they subscribe to is the spreadsheet interface plus dashboards. Write-back is attached to it, and data applications are built from the interface. Neither of these two can collect money separately. The 200 million US dollars is earned by selling dashboards one by one.

Second, what category the industry classifies it into.

Let me first explain what Magic Quadrant is. In the software industry, it is equivalent to the Michelin Guide in the catering industry. Being on the list means that the industry recognizes you.

Gartner has included it in the Magic Quadrant for "Analytics and Business Intelligence Platforms" for two consecutive years, and the quadrant name itself includes BI. It does not count what the company itself claims, the rating agency classifies its business under the BI category.

Its position also has special meaning. It has stayed in the "Niche Players" quadrant for two consecutive years, and has not crossed the threshold of the "Leaders" quadrant for two years. To put it in layman's terms, the industry recognizes that it is a BI vendor, and also recognizes that it focuses on a specific niche, but the premise is still that it is doing BI business.

Third, who are the paying customers.

There is a list of names on the customer page of the official website: food delivery platform DoorDash, private equity giant Blackstone, HR software vendor Workday, and language learning platform Duolingo.

What do these companies buy Sigma for? They use it to make dashboards for order delivery volume, net asset value of funds, and personnel rosters, and present them to the management.

The CEO says customers vote with real money for AI. But after checking the accounts one by one, all the money they pay is for the data viewing service.

After checking the three points, we have to make the ugly truth clear. I checked, and I did not find any cases of customers complaining publicly that the data calculation of Sigma is wrong, or abandoning Sigma and switching back to established tools. Customers are paying their bills very steadily.

Let's take a look at the growth again: in April 2025, the company published a blog post celebrating that its annual revenue exceeded 100 million US dollars. One year later, the official figure became 200 million US dollars, and one month later, the Series E was officially announced. During this period of doubling growth, the main support still comes from dashboards.

So the situation is contradictory. The business that it claims has no future is making money, and the revenue is doubling; the business that is claimed to have a bright big future still has to be supported by the former.

......

Can it afford to support the new business?

I checked, and this year's Series E raised 80 million US dollars. Databricks Ventures joined the list of new shareholders, and Snowflake is an old shareholder from earlier rounds. Data warehouse vendors hold shares in BI companies, and BI companies in turn drive business for data warehouses. This closed loop is very tight, and every party in the loop has no way out.

But the pillar has not been firmly established yet. Sigma Tables, which aims to turn Sigma itself into a database, has not yet come out of beta.

The slogan "Say goodbye to SaaS chaos and build your own applications" is still placed in the showcase for now.

The ranking list will not wait for anyone either. ThoughtSpot entered the "Leaders" quadrant as early as 2019, but Sigma has not got in after two years.

There is another hard threshold. Sigma's gene is that it only runs on cloud data warehouses, and its linkage with Snowflake's data governance does not work outside Snowflake's ecosystem.

The more successful it is, the more inseparable it is from the data warehouse; the stronger the data warehouse is, the more it looks like an interface attached to the data warehouse. It is true that they need each other, but it is also true that it hands over its lifeline to others.

Speaking of the Chinese market, this path is relatively difficult to replicate domestically.

Large domestic customers require on-site deployment, customization, and privatization deployment, and a large amount of data is stored in their own on-premise server rooms. The first premise of Sigma's path, that everyone uses cloud data warehouses, cannot be met in the domestic market.

So this story can only happen in the United States, because the market soil is different.

The meaning of the sentence "We don't want to do BI" should be clear now: the old business is carrying the weight first, the road has not been finished, and the tone has never changed.

At this point, there is one last thing to say. The BI plus AI track is not developing as fast as people imagined; old businesses are often more profitable, and new businesses take time to grow. This is the choice of customers, and customer maturity has not kept up with the pace.

The good thing about Sigma's story is that it is not pretentious. It says it does not want to do BI, but all of its 200 million revenue comes from BI. The old business steadily supports the new business, and the direction has not changed its statement for eight years.

This kind of contradiction is a process that anyone who has done business has experienced, but Sigma just puts it on the table.

Data Reference Sources

[1]. Company and operating data are from Sigma official sources, industry ratings are from Gartner Magic Quadrant reports, Palmer's statement on product direction is from the MAD Podcast interview, and cooperation details between Snowflake and Databricks are from their official announcements

This article is from the WeChat official account "Wang Zhiyuan" (ID