Axioms of Visual Analytics

As I strive to learn more about the visual analytics space, I've tried talking to people in that industry to understand how the industry is structured and what skills are required to participate. On Friday, one of the smart folks over at Juice Analytics generously (and patiently) spent more than an hour with me answering questions about the space. Here are some of the things I've learned:

Axiom 1: Your data elements are not your children - you're not required to love them all equally. This is a famous-ish quote. What it means, essentially, is that just because you have data doesn't mean you're required to use it. Depending on what you're trying to accomplish. some of it might not be necessary. Which leads to...

Axiom 2: You must understand your audience. Executives love targets, trend lines, and accountability matrices. Technicians love navigable detail. But even within these generalizations, there's another level of audience specificity that's required. Who will be looking at your data and why? What will they be wanting to accomplish? Who's presenting the data and what is their agenda? You may have multiple audiences you need to satisfy, each with its own specific needs.

Axiom 3: Act top down, not bottom up. Drive the entire structure from the business decision in question. Don't let the data available determine the question you're going to answer; let the question you need to answer determine the data you need to acquire, even if you don't have it yet. (This is a very common problem in a lot of software applications today, primarily because they evolved up from data acquisition. Five-to-ten years ago, data acquisition was the 'hard part'. Fitting that data into a business context is the hard part now.)

Axiom 4: Be vertical. This is probably a corollary to Axiom 2, but it essentially argues that there is no horizontal solution to a particular analytics problem. You must understand the audience, the business needs, and the data set. What Financial Services needs is going to vary from what Healthcare needs. While there will be common elements, you can't let those blind you to the needs of the market. What this really tells me is that the standard software model of write-once, sell-many is not going to fly here. Of course, that begs the question: If every solution is custom, even within a specific market segment, how do you get rich doing this?

Tomorrow, I'll talk about what I learned about the structure of the industry.