How Companies Evaluate Sigma and ThoughtSpot
ThoughtSpot Search: It's Not What You Think
Imagine you had a search bar where you could type in any question about your business and it would pop out an answer. It would work like Google, use natural language, and live query your data sitting in cloud sources like Snowflake, Databricks, or Redshift. You’d buy that, right? The business would be incredibly productive as they rapidly fired questions all day long using this magical text box. The problem is, it doesn’t exist.
Google search and ChatGPT are testaments to the ingenuity of humans. Search and AI are well-established technologies that bring real value to our everyday lives. They help find answers to almost any question, write poems, and can even drive your car. But, they are miserable when it comes to your own data.
These technologies are fundamentally rooted in the fact that they need to be trained, just like if you were to learn piano or run a marathon. They work great for public domain information on the Internet—which is why Google and ChatGPT are so effective. But, training algorithms to work on your company’s private data is time consuming, expensive, and requires expertise. This is why ThoughtSpot’s promise of search and AI is flat out flawed.
This is also why ThoughtSpot doesn’t use natural language search. It’s not exact enough for business questions. To combat this, ThoughtSpot talks about “keyword search,” which is a euphemism for typing in column names and special keywords. It’s a lot like learning a simplified dialect of SQL.
The search bar has another interesting implication—it’s a dead end. You type in your keywords, get some results, and, well, that’s it. The problem is your business questions are a lot more complex than “how many cat scratchers did I sell last year in Cleveland.” You already know that.
Sigma’s Workflow: The Spreadsheet Way
Business problems are a series of dependent questions. You need to group and filter data, write formulas to express business logic, create cohorts, tie that back to other tables, input some of your own data, filter, and then iterate on this process.
The workflow described above should sound familiar. This is how we use spreadsheets. This is how billions of people worldwide already work with data. It’s a dance with many steps, and you’re not always sure where your partner is going to go next. This is how Sigma works. You work with your data live, similar to how you would with a spreadsheet, but without the limitations of scale. One billion rows of data? No problem.
Finally, there’s the data—your data, sitting in a cloud warehouse like Snowflake. Some of it has been modeled by your data team and is very clean. More likely, a lot of it is less clean or completely raw, straight from the source. Some of your data isn’t even in the warehouse. It’s sitting in Excel or csv files on your laptop. In fact, the most precious of all data isn’t in any of these places. It’s in your head as tribal knowledge. Sigma is built for this world.
To get started with Sigma, all you need is a web browser. You create a Workbook, similar to how you would in Excel. Then, you add your data. This can be from your cloud data warehouse, csv files, copy/paste, or even typed directly in, capturing valuable knowledge from your head into your analysis. You use familiar Excel functions and work with the data like a spreadsheet. Sigma even supports one of the most beloved features of spreadsheets—vlookup, which can tie together all of this data.
Data is never stored in Sigma so governance is straightforward. It all lives in your warehouse and it’s always live. Because it works like a spreadsheet, you no longer need to extract to a spreadsheet. The business will all get the same answers with a single source of truth
Learning Sigma is quick, since you already know spreadsheets. Start dancing with your data and get the answers to real world business problems today. Don’t search, just Sigma.