A well established (Customer Relationship Management) CRM team should comprise of the following components: DBM (Database Marketing), Predictive Analytics, Campaign Measurement. For a large corporation, each of the components will be a team of analysts. For this discussion we will focus into the Predictive Analytics part of CRM for a B2C type business.
The two main focuses the CRM team of the organization would have are customer attrition (Loss of market share) and upsell opportunities (Increase profit). Predictive modelling and customer segmentation are keys to success using available data. The detail on how predictive propensity models are built is out of the scope of this discussion.
Basically the modeller analyzes past customer behaviour and assess key factors that correlate with the variable of interest (what the modeller is trying to predict). Then periodic model scoring will occur to rank score the list (monthly or quarterly) of the customers to assess the likelihood of the variable of interest has changed. The following simple example will let us understand how we can optimize our targeting with predictive modelling.
Suppose company XYZ wants to reach out to 20,000 clients from its 100,000 client base to upsell a new product. Since only 20% of the base is targeted, the marketing team must try to find the 20,000 customers with the highest take rate (probability of buying). The modeller came up with a model for the client base and it effectively labels which 20,000 are best to target with the maximum estimated take rate. The DBM team then use the decile labels and select the 20,000 customers to target. It is not surprising that this type of targeting approach can improve the take rate by 100% (2X) compare to random targeting.