Cindi Howson has a new report out about the importance of interfaces to BI tools.
Cindi does good research, and I don’t disagree with her methods or conclusions. However, it’s worth pointing out a few issues with regard to interfaces for analytical software:
Visual analytics are very appealing, but it’s difficult to visualize complex statistical models using more than a couple of variables—most of us can handle three dimensions at most.
If analytical software is very easy to use, novices may violate a variety of important assumptions about the distribution of data and the limits of the analytical method.
Visual analytics and easy-to-use interfaces are one way to get more people to use analytics, but another way is to embed analytics into transaction systems and business processes so that nobody has to spend a lot of time interacting with the data.
It’s great to have people playing around with data in an interactive fashion, but it would seem that fewer and fewer people have the time for that sort of thing in these trying times.
Interactivity and visual interfaces certainly have their place in BI, where most reporting applications have no statistical meat or underlying assumptions. For analytics, however, the situation is trickier. Most organizations will need to decide just how much they want to focus on the ease of use issue, or whether they will take another approach to making analytical decision-making more pervasive in their organizations.