Speaker: Professor Peter Grindrod, CBE
Location: Gordon Lecture Theatre
Time/Date: 13:00-14:00 on 25th April 2007
Abstract: Service businesses (retail, online, telco, media, finance,...) recognize that their most important assets are their customers, and their goals must include growing customer value. We consider the practical problem of modelling and forecasting customers’ bahaviour changes that give rise to value changes. This is particularly challenging when the customer base is vast with tens of millions of individuals all transacting. We introduce a class of dynamic (hidden) Markov chain models suitable for businesses with large customers bases, where individuals may transact regularly in different ways, and their behaviour over some time period is summarized by a number of distinct metrics. Such models, once calibrated yield calculations of customer lifetime of horizon value that can be used to identify opportunities to grow value, and forecast the impact of proposed activities. We discuss the user options of such models, their calibration using genetic algorithms, and possible optimization to enhance their forecasting and operational usefulness. We show how these concepts and the derived results are relevant to both management and marketing operations, and we illustrate these methods with three real case studies. We also discuss how such models can be deployed in practice within the IT infrastructure of a typical user.
Valid XHTML | Copyright | Last Modified: 26/Apr/2007 |