|
Jeffery S. Saltz Vice President, J.P. Morgan | Jonathan M. Steinbach Associate J.P. Morgan |
Introduction:
There are two key problems to overcome when creating a useful information visualization. The first task is to define the visualization; which relies heavily upon brainstorming and innovation. However, the innovative display of the information, by itself, is not enough to ensure the success of the information visualization. In particular, users must intuitively understand the visualization that has been created. Most papers in the field of information visualization concern themselves with the innovative display of information, but less research is done to make sense that these visualizations allow users to obtain a better (and/or quicker) understanding of the information presented.
Information visualization at J.P. Morgan:
In financial visualization, the date. Is often numerical. However, unlike the scientific data visualization, there is no inherent geometry in this information. No matter how one rotates a car showing stress analysis, it is easy to see where the fender is located. This is not so easy when one is viewing a risk assessment.
At J.P. Morgan, we have half-a-dozen people creating information visualizations. As we have very little "visualization research" funding, our users pay for all their visualizations. As one might expect, this creates motivated users who are willing to work with us to create something useful. It also separates the useful from the unusual visualization visualizations, since those which are beneficial receive funding.
When we create a visualizations at J.P. Morgan, it is a very interactive and iterative process involving our end users. This is a key part of our ability to create useful visualizations. Without testing our ideas on target users, and without getting ideas from those users, our visualizations would look much different, and more importantly, would be a lot of less value to our clients (i.e. end users). Intel Corp. group, approximately half of the final information and visualizations have been proposed by an users during a brainstorming session with us. In general, it would be interesting to know how many information visualizations have been proposed by end users?
Selecting appropriate visualizations:
The are many interesting questions that can, and should be asked. For example:
At J.P. Morgan, these usability questions are asked all the time. After all, who wants to pay extra for a tool that could just as easily be solved with a spreadsheet. We often cannot fully answer these questions until prototypes are built, but from a user's perspective, these questions all get rolled up into "can you create a useful visualizations for me" or even "this is the problem I am trying to solve, can visualization help me. "
For example, there has been much discussion of 3D visualizations verses 2D visualizations. It would be interesting to see if the 3D screens were truly better (or even a perceived better by users). Within J.P. Morgan, it has become clear that some users do not want to see 3D, even if we think it will enable the creation of a much better visualization. Others want 3D no matter what we think. Again, 3D verses 2D is not the point. The idea that both the creation and understanding of the visualization is important.
It creating information visualizations, we always try to extend a metaphor that our users understand or that fits well into their current working environment. This is typically not a physical metaphor such as a room, but rather the line charts, bar graphs and spreadsheets which are commonplace in the financial world. In the end, most of our users think "how does this relate to my 50 spreadsheets" (some of which have simple graphs). When we have created visualizations that were not based on a metaphor that users can understand, the visualizations have not been a success.
Summary:
It would be interesting to see more research in the area of how people understand the visualizations that are being created. We have found that many of our users prefer visualizations that display less information more intuitively, rather than visualizations that display more information, but in a less intuitive visualization. Many times, there is a trade off between simple visualizations and more informative visualizations. Our work with end-users suggests that users must quickly understand the visualization. So, if we create very informative visualizations that take some time to get used to (or worse, take a long time to understand even after we are used to the landscape), then, often these visualizations fail to be useful from a user's perspective.