Over the past year, my column has argued repeatedly that lifetime value is the one true measure of business results. But lifetime value deals with customers as a group. The central question it answers is, how much value will this set of customers produce? This tells how much the company can pay to acquire those customers or keep them from leaving. Components of lifetime value, such as orders per customer or trends in the retention rates, also provide valuable insights into the health of the business as a whole.

However, sometimes you need to look at customers as individuals. In fact, most day-to-day customer management decisions are made at this level. Which offer should I send? Where should I set the credit limit? How should I respond to this complaint? The criteria for resolving those questions still impact lifetime value, but lifetime value analysis itself cannot predict how particular customers will react. This requires a more precise tool that looks at the behavior of individual customers.

Such behavior analysis is aimed at identifying differences among customers that predict their future activities. This focus on differences rather than the group is what distinguishes behavior analysis from lifetime value analysis. In other words, behavior analysis is really segmentation. It encompasses simple recency-frequency-monetary (RFM) value cells, predictive modeling based on statistical techniques and specialized behavior monitoring systems that identify patterns and deviations from patterns in customer transactions. The success of any approach is measured by how accurately it distinguishes customers with different future behaviors.

Most behavior analysis is aimed at estimating the impact of taking a particular action, such as sending a marketing message or applying a particular customer service policy. But behavior analysis can also measure what follows an activity initiated by the customer, such as buying a new product or canceling an old one, or the results of events that neither party chose, such as a snowstorm or product failure. It sometimes predicts behavior of customers absent any particular event, for example, as part of a customer value inventory.

Behavior analysis and lifetime value analysis both start with transactions and nontransactional attributes (such as location and demographics) that have been linked to individual customers. Lifetime value analysis aggregates this information almost immediately, while behavior analysis continues to work with the details. Because different details will be important for predicting different events, a behavior analysis system requires a repository that keeps all the detailed data and allows analysts to extract the relevant bits as needed for a particular project.

More data will be extracted for analysis than is ultimately used to make a prediction. One purpose of the analysis project is to determine which elements are actually needed for a particular goal. Only this data, or a summarized result such as an RFM code or model score, is then exposed to the operational systems. The less data moved to the operational system, the more efficiently everything will run. However, data gathered by the operational system must be combined with pre-existing details to make a prediction; there may be no choice but to export the details themselves.

Integration with operational systems imposes other technical requirements as well. These include interpreting customer activities that occur during the interaction; access to detailed historical data and/or appropriate summaries; integration with rules engines, content management, offer management, inventory and other systems; presentation of results back to the operational system conducting the interaction; and failing gracefully in the face of unexpected conditions. An effective behavior analysis infrastructure must also capture new events so it can find changes in behavior patterns as these evolve. Re-analysis of behavior need not occur in real time - significant patterns don't change that quickly, and you probably want some human oversight -but it does need to be frequent.

It is easy to contrast behavior analysis and lifetime value, but they are really more complementary than competitive. It is possible to model an entire lifetime of interactions at the individual customer level and thereby derive a lifetime value estimate from behavior analysis alone. You can even buy software built for this very purpose. However, it is more common to use behavior analysis to identify segments at a single decision point in a customer relationship, test different treatments within each segment and use lifetime value analysis to measure the results. This does not provide quite as much insight into the behavior of the different segments as analyzing their subsequent activities in detail, but it does tell which approach will yield a higher lifetime value. And lifetime value is the one true measure of business results. 

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access