SM: How long did it take you to build and flip BeeLine, and what did you do after that?
CC: It took about 18 months. Even to this day I’ll look over at a stoplight and see one of our maps, which is very satisfying. A month after we sold BeeLine we had a party to celebrate, and I ran into a friend from business school. He told me about an opening at a venture capitalist firm he was with and convinced me to join him at this firm, which was called Softbank Venture Capital. It had a notorious reputation, but despite its woes had many successes. I did that for two years, and then I got the entrepreneur itch again. I wanted to start another company and this time I had no interest in flipping it.
SM: Did you come up with the idea for your next company while you were working at the venture capital firm?
CC: No. The idea came out of relationships. My first company I founded with two computer scientists named Christopher Stolte and Maneesh Agrawala. We started meeting and talking about ideas for the next company, and we were very patient to make sure that we found the right idea. Ultimately we decided to commercialize an idea at Stanford called the Polaris Project. That is what became Tableau. This all occurred around 2003.
SM: What was the idea behind the Polaris Project?
CC: The idea was to make database structured data easy to visualize and explore.
SM: What was the state of the art regarding database visualization at the time, and why was Polaris different?
CC: Almost all visualization of data, even today, follows the same archaic model. First you open some data with a query interface and you work with that data. You analyze it, dice it, and pivot it, all in text form until you get what you would call your answer. Only then do you put it into some kind of chart wizard. Once you get your data points into the chart you have an end result, which is data translation. And what happens next? You look at it and say, “That’s not what I wanted” or, “That’s what I wanted.”
Your brain is naturally curious about data whenever it sees it. The problem with the whole paradigm to understanding data is that the visualization comes last. By then it’s too late. If you have a new hunch or angle, then you have to go back and do the whole process again. The idea behind Polaris was to query a database using a picture, to be able to sort, filter, zoom up, and pivot it through a purely graphical interface. When you do it that way you are working at the speed of thought. By dragging and dropping after viewing some of the data on a canvas, you are actively querying it. That lets you generate pictures of it at the same time.
SM: Architecturally, how does the data tie to the graphics or picture?
CC: The core invention of Tableau, which is what it will be known for, is VizQL. One of the most important advances in using data was SQL. The idea behind SQL was to have a pithy declaration that was almost plain English to send to a database and let the database find the answer. It was declarative, not procedural, and it changed the world.
Regardless of what you send, SQL always replies with a table. You then take that table and go through the clumsy process that I just described above to get to an answer or presentation. The idea behind VizQL is to be able to send a VizQL statement to a database and have it reply with a picture, not a table. You would just turn that table into a picture of some sort anyways. VizQL is an algebraic formalism that embodies both the graphics commands and the query which is required to bring tuples back into the data engine. By virtue of marrying both into a single language, it is easy to provide a single picture of the data.