What is the best way to approach visualization projects?


What is the best way to approach visualization projects?

I work on several areas related to information visualization, linked data, computer vision and other stuff, so mainly front-end. I am not really happy with the fact that visualizations take lots of iterations and lots of time from first prototype to production ready code. So here is the question: what can I do to improve that?
Sometimes we do a prototype in few hours, add something, do another prototype, rewrite it to use jQuery instead of Prototype because it's a better fit and so on and so forth, but almost all visualizations go trough 15-20 iterations (sometimes taking up to few months of development time - the code will also go from the 300 lines of the initial prototype to 2000 or 3000 lines due to evolving requirements and iterations). Since I am working at multiple projects, from my point of view the fact that sometimes it can take months is not a problem, just that I am not really happy, as I could do 40 visualizations per year instead of 15 for example, if I would only be able to improve this process by reducing the number of iterations or by any other means.


Answer 1:

Information is beautiful is a good site to keep reading. It has some truly amazing visualisation of data.

What is vital is to understand what the end users wants and how to display it in an informative way. Using such tools as R, you can mock up some graphs easily and fast to make sure that your users know what they are getting. Having a portfolio of visualisation does help as well as this can serve as a basis for your new project — in the same way as the above link can.

Finally, start (really) simple and add complexity later on. This does mean you will have to refactor large chunk of your code.

From comment: How do you know at which iteration to stop, provided that your code does not have serious bugs?

  • The mercenary approach: when the money the client gave you runs out.
  • The academic approach: when you can write a paper about it.
  • The pragmatic approach: when all your other projects have trumped this one.
  • The perfectionist approach: My children’s children’s children will still be working on this.

For myself, it is when the visualisation tells me things I did not know or expect about my data. Once that is done, I either understand the data (thus write a paper) or the data is incomplete in some sense (new hypothesis time).