Liability and GIS

A great session here at the ESRI Business GeoInfo Summit tackled the use of spatial data to mitigate liability. The first paper by Michael Sullivan of the Liability Management Systems addressed using neural networks to predict how much the manufacturer of a building product (he could not say what) would be likely to pay per census track based on product failures. While in the end he could not tell why the products failed, he could model very well which census tracks were likely to cost money.

Graham Duthie from Anderson (Windows) Corporation followed up with not a story of warranties of windows, but warranties of beach balls. He can’t really talk about windows… The problem was that return rates of beach balls seemed to up in certain areas. Were they really or was it an anomaly? He used some basic state (chi squared, which I’d not seen since college) to tease out the pattern and determine UV activity as the principle issue for failures around the equator.

These seem like straightforward ways to help companies make money/save money – ideally before too many failures. Unfortunately, these ideas, so the experts tell, are not taught in business school and are not widely used in commercial companies.