Customer retention remains a top priority for most large consumer businesses. Yet the standard approaches to targeting customers "to be saved" often perform badly, with retention activity frequently back-firing for a minority of customers, and occasionally actually driving away more in total than it saves.
How can this be, and what can be done about it?
Many companies engaged in retention activity have suspected or measured negative impacts for some segments of customers. Undesirable as they are, it's not hard to understand how these negative effects arise.
The typical state-of-the art approach to retention contains several key elements. First, the company will identify customers at risk of leaving. This will probably be based on a churn model (or an attrition model), which predicts how likely customers are to leave. In other cases, all customers coming to the end of a contract period or some other kind of "lock-in" will be assumed to be attrition risks. It would also be normal to take into account customer value, most commonly with some sort of historical measure, or sometimes with a projected value over some period. High-value customers judged to be at high risk of leaving are then targeted. The retention activity itself can take many forms, from a customer care phone call, perhaps with an offer, to some kind of mailing (paper or electronic), or a communication through a statement or the web. Alternatively, a customer may simply be flagged for special attention when he or she next contacts the company.
Though sometimes effective, there are number of problems with this approach.
Stochastic Solutions staff have world-leading experience in this customer retention modelling, and can help guide companies towards the most effective, modern approaches to retaining more customers. This includes evaluating the impact of past campaigns, where possible, designing new campaigns whose impact can be measured accurately if not, and preparing and helping the company through an evaluation and possible adoption of uplift modelling (incremental modelling) as a provably more effective, more profitable and usually cheaper way of increasing customer retention. The fundamental change is from targeting customers based on their probability of leaving, to targeting customers who can actually be saved.
This approach allows the customer base to be segmented in the ideal way, enabling companies to see which of their customers are likely to react positively to a given intervention, which are likely to be unaffected by it, and which are likely to be adversely affected, leading to increased attrition. This is illustrated below. Notice how, when there are negative effects, optimal targeting not allows only fewer people to be targeted, but actually saves more customers than does targeting everyone. In the illustration below, targeting 40% of the base actually saves 10% more customers than does targeting them all.
It is also possible to extend this approach to the case where multiple treatments are available, and the problem becomes matching the best treatment to each customer.
