Empirical Testing of Commercial Visualization Tools
===================================================

Solved three problems: frequency of cheating among guys and girls;
technical data about cars; location of heavy metals in Sweden.

Significantly faster (about 80 sec) with InfoZoom than with
Eureka or Spotfire (both about 105 sec).

Eureka
------

Users confused by hidden variable labels (abbreviated to one or two letters).

Difficulties with three or more attributes (problems remembering the colour
code, problems finding the right filtering and grouping strategies).

Problems seeing correlations.

InfoZoom
--------

Problems seeing correlations.

Easy to expand lines, but people didn't do it.

Zooming and backtracking were a very natural and powerful operation.

SpotFire
--------

High cognitive setup costs: took considerable time to decide on the right
representation and correctly set all the parameters.  Caused by the wealth
of different visualizations, but also by the restrictions each of them
imposed once it had been selected.

Often used scatterplot first, because it was default.

Users would try to stick to their first choice of visualization, trying
to adapt the visualization to their needs instead of going back and
choosing a better visualization.

Success Factors
---------------

Practically no differences observed between 2D display of geographic
information in SpotFire and 1D display of X-Y values in SpotFire and
Eureka!

1. Properties of a visualization: use text, or at least make it easily
accessible.  Assist usersin deciding how to map a problem into the
representation.

2. Operations that can be performed: zooming in InfoZoom vs. mostly
context-preserving operations in Eureka.

3. Concrete implementation.  Eureka and InfoZoom both offer a table-like
visualization.  But the view is rotated by 90 degrees -- this has a big
impact on variable labels and recognition of correlations.

4. Consider visualization-independent usability factors.  ("Priming"
in SpotFire.)