Interactive Visualization of High Dimensional Marketing Data in the Financial Industry

Abstract

We describe the visualization of high dimensional marketing data for a financial asset management company. The data typically consists of 30 to a 100 variables of 25000 to half a million clients. We use the visualization of the correlation matrix as a variable selection tool which makes it easier to find patterns in the data. The user can then select data ranges of the selected variables and start a cluster analysis using 5 variables. The clustered data are then visualized as a set of spheres. In an additional visualization we first sort data values of a client variable and then visualize the sorted cubic data in a cube using volume rendering and isosurfaces. The interactive correlation visualization allows marketing researchers to quickly explore all kinds of combinations of variables, which enables them to find valuable client behavior patterns much faster. The cluster visualization allowed researchers to identify detailed groups of customer with similar behavior. Additionally, the visualization of the sorted cubic data gives in dept information of one variable over the total sample of customers. With these visualizations, a better understanding is given on customer behavior.