Skip to main content
Customer Segmentation to increase profits

Customer segmentation is often the first kind of analytics attempted by a company. It is not, however, only for "newbies" - even the most sophisticated companies can use better segmentation to drive better results.

Take Barclaycard, one of Europe's largest credit card companies and a very sophisticated company when it comes to analytics. They had seen consistently good results from their account management strategies, with “champion” strategies performing solidly. How then to do better? Their approach was to work on developing a better customer segmentation approach and then using more finely defined customer segments to better target the customers in each segment.

They used a CART-based tool called Model Builder for Decision Trees from Fair Isaac , a tool that uses historical data to rapidly create new decision trees to segment customers using multiple performance dimensions. This ability to segment first, say, on retention and then on profitability and then on likelyhood to respond (rather than having to use a single perofrmance dimension throughout the tree) is a key advantage of this implementation of CART. Model Builder also let them process large datasets rapidly (key to such a large organization) so new results can be seen in minutes and handles selecting the most predictive variables for any given outcome and defining the best splits (into multiple tree branches) based on these.

Using the tool they could develop new trees, profile them for greater understanding, compare different approaches to see which customers would be swapped between segments with an alternative approach and, utlimately, transfer a new strategy to production with ease, putting new strategies into production faster.

“The strategies built using Model Builder use new decision keys, have finer segmentation where this adds most value and have tree splits that are closer to being optimal, given
our goals,”

Barclaycard uses Fair Isaac’s TRIAD adaptive control system to manage its 11 million cards - the name comes from the combination of three elements (data, rules, analytics) and the approach is one of "adaptive control" as the software is designed to let you continually adapt your control of the strategies you adopt to manage cusotmers and accounts. Having an EDM platform like TRIAD to implement new customer segmentation dramatically increases the value of that segmentation and reduces implementation time - Barclaycard took only three months from starting analysis to being in operation with the new, tested approach.

You can find a detailed case study here.

related posts