In the last few weeks, there have been a couple of acquisitions in the visualization space. Spotfire (by TIBCO) and InXight by BusinessObjects (the latter primarily for the unstructured text capabilities, but inXight has been known for its visualization too). In each, it is clear that the acquirers believe that visualization of information that is controlled by them in their clients' shops would be an added differentiator.
Of course, any "information" whose sole purpose is to not just feed "processes", and is to be consumed by humans, benefits from a visual interface. I want to separate out custom visualizations -- done with careful, editorial thought, for possibly generations to admire (remember the Napolean's Russian Invasion Fiasco visualization by Charles Minard?). They are a league of their own, and meant for posters and expensive coffee table books (Tufte has many more examples).
I want to focus on what enterprises will pay for in terms of software. In my mind, the challenges here are two-fold. One, to date, most visual techniques work well on hundreds or thousands of rows, not on the millions or billions that typically might be found in enterprises. Consequently, visualizations can convey information succintly (i.e. the interesting facts are known a priori), but can they really help in the determining the interesting patterns? I am sure there is a lot of work here, but I would like to highlight some work from Martin Wattenberg and his colleagues at IBM's Visual Communications Lab on the OLAP of visualization... And when this nut is cracked, enterprises will stand up and notice as visualization not being downstream from information discovery, but a critical component of information discovery.
The second challenge is that visualization techniques depend on very clean data. Some would say, so does OLAP. True. However, in my view, it is even more complicated than traditional business intelligence. As my colleague, Michelle Zhou points out in a soon-to-be-published paper, the transformations needed for visualization are best if they are "data" driven and determined at run time, as opposed to decided statically at design time.
If visualization is to make a signficant dent in how information is consumed in businesses (outside the excel spreadsheet, dashboard stuff, of course), then it will have to be much easier to produce and consume both the information, and the visualiations associated with it. Many people are experimenting around it. I like the Many Eyes experiment around social aspects of visualization, and we will be combining our Info 2.0 technologies with it so that information can be manipulated before and after the visualization step. But others will attempt other techniques.
From a research agenda perspective, my other colleagues Laura Haas gave a keynote at ICDE 2007 on why the HCI community and the IM community need to come together to jointly solve the information and interaction problems, and how, when communities have come together in the past, fundamental breakthroughs have happened. A broader vision than what I have outlined here, but along the same lines.
PS: Some of the links to papers mentioned here are missing. I am going to post them in a new blog entry as they become available, or if sooner, in an edit of this blog entry.