Skip to main content
Engaging Fickle Customers and Tackling Silent Attrition

At the Credit Scoring and Credit Control XIV conference August 26-28, I will be discussing innovative analytic techniques and applications revolving around engaging fickle customers and tackling silent attrition.  Here is a preview of my talk:

Everyone knows the ideal customer: frequent shopper, highly profitable, engaged, loyal. But in today’s competitive markets the most profitable customers are a fickle species and can turn away (or more likely are lured away) on a dime, possibly without giving prior warnings. “Silent attrition” happens when customers stop transacting without saying “I’m no longer your customer”. It occurs in non-contractual relationships such as in retail and it happens with credit cards when customers stop using a card without canceling it.

When silent attrition emerges, businesses have a brief window of opportunity to try to re-engage the customer before the parting becomes cemented. Rapid detection of silent attrition and fast contact through mobile channels provides an interesting option to defend valuable customers at the crossroads, if the business is also able to create persuasive offers.

FICO Labs has been re-thinking attrition analytics and customer dialogue with special emphasis on nimble detection and intelligent action. Unlike traditional attrition models used by many card marketers where scores are updated monthly from Master file variables, our ultra-dynamic attrition models are updated daily, by exploiting granular transaction information. We leverage machine learning that captures complex predictive patterns from massive data and turns these patterns into accurate daily attrition risk estimates. The algorithms can also predict category-level attrition, such as when a customer stops using a card at service stations while continuing other uses.

Decision analytics is responsible for creating effective actions and persuasive offers. The challenge is to move beyond one-size-fits-all retention offers to customized offers that appeal to individual preferences while keeping costs in check. Traditional marketing campaigns are often designed using “single shot” decision analytics. For example a retention decision is based on a snapshot of: attrition score value, last year’s spending, and possibly some other customer profile variables or segment membership. Past behavior informs but can also limit our understanding of preferences, which may have recently changed and thereby possibly created the attrition problem.

To understand today’s preferences and avoid customer fatigue businesses can:

  • Leverage investments in mobile channels and apps to entertain customer dialogues. For example, a customer dialogue could be triggered when attrition risk exceeds a threshold.
  • Ask a simple question or two – whether the customer would like a new card feature, or whether they prefer higher rewards over lower fees, etc. A fraction of the marketing budget could be allocated to incentivize response.
  • Make a more relevant offer to persuade the customer based on the customer’s response (or non-response). Here we reason about a decision sequence – initial decisions concern whether/what question(s) to ask, and a subsequent decision concerns the offer based on new response information that is unavailable for a single shot decision.
  • Perform sequential decision analysis to show positive ROI of such a system subject to assumptions on imperfect dialogue response and incentive costs.
Sequential decision analytics is more widely applicable and valuable. As the old adage goes, 20 percent of customers generate 80 percent of sales. The other customers are lackluster in various shades. Quite possibly a substantial proportion of customers haven’t had a single transaction for months or years – our silent attriters. But their marketing and operational costs keep mounting. In the case of credit cards they constitute contingent liabilities. Businesses may be tempted to curtail such stale customer relationships. The question is how long a business should try to re-engage its silent attriters before cutting costs? And should this decision depend on characteristics of the earlier active relationship with the customer? Sequential decision analytics with the objective to maximize customer lifetime value provides a formal approach to tackle these questions and can contribute substantial business value.

related posts