In part two of our Q&A with Tableau Co-Founder Pat Hanrahan, we discuss the lack of capabilities in most data visualization technologies; the different ways to judge data visualization and the technology developments underway to visualize unstructured data.

Here’s an excerpt from the interview.

Question: What are the goofier forms of data visualization?

Things that are not clear what they are useful for. They look pretty but you are not sure what they are use used for. That is cool but I would not use that every day. Or another way to look at it: Is it useful for one or two things.

That’s the problem – visualization has been associated with being used for these weird things. A lot of people using Tableau – they just want to get their work done. They want a map. With few capabilities, it forces people to use different technologies with different capabilities. The maps may not be hooked up to your data warehouse. People have to use Excel for tables and something else for graphs. That causes people to switch context.

To me one of the big differences is what can you do with visualization. There are visualizations that you can just look at. There are visualizations that you can change or interact with. And that’s an enormous difference. Because if you just look at it – it is what it is. To me, I gave a few talks about this recently, what is neat about math and programming is you can manipulate math. You re-arrange the equations. When you re-arrange equations you prove theorems and solve problems. I don’t like to read programs – well I do – but I like to write programs. When I write them I get creative and I get to build technology. So visualizations are not in the right stage very often.
They are sitting there static entities telling people what you saw.

The basic thing about Tableau, the greatest innovation, we prove that we can formulate any database technology by composing a picture of the answer. That means it is more like equation – a little bit.

Question: Eliminating the complexity means you can optimize it. Right?

A lot of visualization is about widgets. Bar charts, pie charts. Instead of query languages perhaps it had pre-canned queries. I bought this sum of sales app, I bought the market analysis app

What we did was build VizQL. We created a language for making visualization. We create a huge variety of visualization. If you look at our papers it is all about composable, adaptable technology. It is a lot more useful in different context.

Going back to you questions about the different types of visualizations, what are some of the ways of judging visualizations.

1: What type of visualizations?
2: What can you do with it?
3: How composable and flexible is it?

Most visualization technology does not attempt to go into the two or three camps. That is unless you have a programmer involved. Most people are not programmers.

Question: So they need to know the simple steps.  To make changes to the data so they can see the visualization that comes of it?

Exactly. A good example is Tableau Public where you can upload visualizations to the cloud. Our audience for that is a journalist. He knows Excel, Power Point, Office type things or Apple. He knows the office technology. He uses that to produce content. They should be able to produce visualization using data and other tools. They should be able to upload it and then click to get a URL back. It should be as easy to make a visualization for an author as it is to make a blog. We are a long way from that.

Question: The amount of unstructured data we ar seeing now is tremendous. What technology development do you have underway to turn that dat into visualization?

That area of visualization needs a lot more research. There have been a lot of neat, little attempts at that.  What I am seeing some success is in semi-structured text analytics.

You do one of two things. You take a bunch of articles, text, email documents. You think of each of those as a virtual record in a database and you give it a bunch of properties. You create little smart extractors. In email, you can extract the “from” and the “to” or who it about, when did it occur. At least it is presented to you as a kind of table and then you can start leveraging it. You can put all the articles on a map and do things like that. It’s not ideal but it is one approach.

The other approach is to take more like a table like approach and add text to that. Say someone is filling a form and there is a field for notes. And the notes are random text. You have some structure. You know it is an application or the record of some event happening. You have some text associated with it you’d like to analyze. You have some sort of structured content.

Free form visualizing text – there has been some neat visualizations. But none of them seem like a tool yet. They are more like experiments. It is sort of a neat picture someone took of that text.

Question: It almost seems as if has to be done by some spatial, time perspective?

That is where there is a lot of progress. People register things in space and time coordinates. Google Finance – you have an event in the stock market, you figure out what articles were published then. You see similar things in mapping.

Another way to look at it is there is a class of data and somehow we have to get the data in there. Like you take an article and put it on to a map. And it tries as hard as it can to see where it goes. If it can’t figure it out, it will pop up a little box and ask where it should go.

That is a wide open area (unstructured data and visualization). I follow that area quite a bit. But it still has a long way to go.

Services Angle

I want to thank Pat Hanrahan for taking time to talk with me during his visit to Portland. My takeaway – data visualization will force us as a society to look at information in a manner that will make long, text-based documents seem antiquated. We can’t understand huge data sets with words. We will increasingly need pictures that tell a story. That shift will happen through the advancements in services from companies such as Tableau Software.