The most practical way to build knowledge on customer behavior is to produce scores that explain a target variable, such as churn, appetency, or upselling. The score is computed by a model using input variables that describe customers; for example, their current subscription, purchased devices, consumed minutes, and so on. The scores are then used by the information system for things like providing relevant personalized marketing actions.

A customer is the main entity in most of the customer-based relationship databases; getting to know the customer's behavior is important. The customer's behavior produces a score in relation to the churn, appetency, or upselling. The basic idea is to produce a score using a computational model, which may use different parameters, such as the current subscription of the customer, devices purchased, minutes consumed, and so on. Once the score is formed, it is used by the information system to decide on the next strategy, which is especially designed for the customer, based on his or her behavior.

In 2009, the conference on KDD organized a machine learning challenge on customer relationship prediction.