Outra power of change in segmentation

The power of change in segmentation

It is often said that the only constant is change. Why, therefore is change rarely used as a variable in segmentation? After all the purpose of segmentation is to create powerful models that make predictions based upon the data that feed them.

For instance the surprise Brexit vote could easily have been predicted if changes in immigration patterns had been analysed before the referendum. The areas in the UK that had experienced a significant net increase in immigration since 2011 were more likely to vote Remain, whilst locations which over the course of six years remained stable in terms of their multiculturialism voted Leave. In this case change was the predicative variable.

So how can this be applied to marketing?

Very easily. For instance in the case of consumer credit. Typically credit providers use scorecards based on huge volumes of demographic and financial behaviour data to determine which customers should be offered extended levels of credit.

But these models, in our experience, are seldom as predictive as they should be. And the reason for this is that they do not use longitudinal transactional data to track changes in the financial situation of customers.

The scorecards tend to feed a single segmentation which is credit worthiness rather than a two dimensional segmentation which should also contain financial transaction data over periods of time. If change was made the predictive variable more appropriate limits could be offered to customers making the credit industry more responsible.

Over the past 30 years the trend in the data industry has been for increasingly granular data to be used as the foundations for classification systems, not least as a result of improved computing power.

As this has happened, the misconception that more detailed data is better than less granular data has manifested itself. However, often this is not the case.

Sometimes the most powerful data is tracking change rather than volume or depth of the data.

So the question to ask yourself is what changes can you track and would they form the basis of a powerful prediction for your business?