Barclays Moves Into The Predictive Modelling Fast Lane

A quiet revolution is already taking place behind the scenes in the bank's marketing department where new generation business analytics technology is helping sharpen the marketing of new financial products.

While analytics in financial marketing is nothing new, Barclays is one of the first to embrace the radically different science of mathematical rather than statistical modelling to build the predictive models around which analytics is based.

The result has been a significant increase in speed, productivity, flexibility and ease of use, giving Barclays a new edge in marketing its products with tangible income benefits. Thanks to its chosen analytics solution - the KXEN Analytic Framework - the bank can now create many models every month.

"Previously, with a team of dedicated modellers, it would take anything from four to six weeks to build a single model," says Barclay's senior modelling analyst Robert Grimm. "But with KXEN and a few other improvements we're down to one or two weeks with just one full time modeller."

The productivity gains mean Barclays can create many more models than before, including quick one-off models and speculative 'what if' models, neither of which were really an option before. Being able to build and test multiple models early on in a campaign ultimately leads to more accuracy, believes Barclays.

"Now we don't have to put so much time into the model building process it gives us the option to test out new ideas and be more creative. It also means we can build quick, on-demand models for one-off campaigns and, even if we don't use that model ever again, it's not a problem," says Robert Grimm.

Behind the KXEN software's advances are its basis in breakthrough mathematical thinking by the Russian mathematician Vladimir Vapnik. Unlike conventional statistics-based tools, it allows models to be built in just a few hours rather than several weeks or months. Marketing staff can have the answers to business questions much faster, critical in getting new products to market ahead of the competition.

Another benefit of KXEN is the number of variables it works with. Conventional modelling requires analysts to restrict variables to a few tens or hundreds. KXEN has no such limit and routinely generates models accepting thousands of variables. In practice this lets users consider many more factors about a product and its target audience, letting them arrive at a much more informed conclusion about who might buy it.

This hundreds-of-models, thousands-of-variables philosophy has been dubbed extreme modelling by KXEN. It is important to the way Barclays uses the product, as Robert Grimm explains.

"Without KXEN we could look at maybe a few 100 variables at most, now we can consider several thousand. This allows us to factor in all sorts of data that we couldn't use before and it might well be that just one of those extra variables is the one that is key to a whole campaign."

The additional scope for variables lies behind much of the increased productivity. Barclays' Teradata warehouse holds information on some 14 million active customers with data on many different aspects of customer behaviour. Previously, analysts had to manually whittle down several thousand fields of data to a core of only a few hundred to fit the limitations of the modelling process. Now all of the variables can be fed straight into the predictive model.

"KXEN has taken a lot of the hard labour out of that initial variable selection. It used to be a fairly manual process that took as long as four to six weeks," says Robert Grimm. "Now KXEN does the whole job itself of deciding which variables are going to be predictive. Indeed that fact alone was of the main reasons we chose KXEN."

Other factors that had a bearing on Barclay's decision were KXEN's speed, its capacity to deal with large datasets and its ability to read in SAS data directly. Also critical were ease of use and the way in which KXEN outputs code that can be immediately applied to scoring models on the database.

At Barclays the work with KXEN is concentrated in the retail area where it not only informs the targeting of sales campaigns but also the most effective channel over which to reach customers and prospects. The recent purchase of further KXEN licences points the way forward for use in other areas at the Bank.

"Accurately targeted marketing is important in banking and it's all about making our campaigns more efficient and more profitable," says Robert Grimm "Modelling is a crucial part of that process and with KXEN we can build more models to apply to more campaigns, and get better effectiveness as a result."
www.kxen.com/cxo