A background in database design is an enormous advantage in any data analysis. The underlying structure of a dataset provides an ideal framework in which to identify patterns of behaviour. The normalisation and design processes use many of the same statistical techniques that an analyst would to identify the characteristics of data.

 

 

 

The use of simple constructs like cross-tabulation, cluster or factor analysis are the realistic limit of our ambition so far and I remain nervous about some of the techniques used to forecast value, based as they are upon the foundations of very limited data.