The buzzwords of the moment in business intelligence seem to be predictive analytics. These words encompass two notions: all the data we've been gathering should be used to predict customer actions (in addition to reporting on them), and these predictions should be used in operational systems to help guide interactions. In other words, we need headlights in addition to a dashboard, and the headlights should be placed so the driver can see what they show.

This makes sense, but driving safely is one thing, and knowing where to go is quite something else. Most predictive analytic applications are strictly tactical: which offer is the customer more likely to accept or, at best, which will yield the most profit? Yet maximizing profit on a single interaction may actually reduce the value of the long-term relationship. An inappropriate offer might annoy many customers or preempt another offer with less immediate return but greater ultimate value.

Good businesspeople understand this intuitively. It's why they offer special discounts to retain new customers and bend the rules when their most profitable customers have a problem. The challenge is translating this intuition into corporate policies that optimize results.

Such policies must be based on empirical analysis, but the proper data is rarely available. Unless a formal test of alternative treatments is in place, companies treat all customers according to the same business rules. Thus, the best an analyst can do is look at how customers behaved after being given certain treatments in certain situations.

This may identify obvious problems, such as high attrition rates when an offer is made following a service problem. The inference is that something other than an offer would reduce attrition in that situation. However, it's a tenuous link at best because the data can't show what would have happened had the company done something else. Only a formal test - randomly making different offers to customers in similar situations and tracking their subsequent behavior - can really isolate the effect of that single decision. Such tests are often difficult to set up and take a long time to evaluate. Managers may also be reluctant to treat some customers differently than others, fearing that some will feel discriminated against.

These difficulties should not be overstated. In practice, some interactions are clearly more important than others. First-time buyers, customers who have had problems and customers approaching a renewal date are at obvious inflection points. Companies can test alternative treatments in these situations with reasonable assurance that differences in near-term results will correlate with differences in long-term value. Once the obvious candidates have been optimized, the company can work on more subtle options such as changes in contact frequency or different loyalty incentives.

If marketing treatments were the only choices that companies had to make, offer testing would suffice to optimize results. However, customers are affected by many things: product design, manufacturing processes, logistics, customer service and even personnel policies. Many of these decisions compete for financial resources; all compete for management attention. Each needs to be assessed in terms of its impact on customer relationships. Although it's a bit of a cliche, the aggregate value of these relationships truly is the value of the company itself. Thus, impact on customer value can be used as a standard metric to compare investment opportunities in all business areas.

Why use customer value instead of a traditional measure such as return on investment? The reason is that calculating customer value requires modeling the major interactions between the company and its customers; that is, it forces the company to simulate the customer experience. This leads to an understanding of the interconnections of different business decisions, which traditional financial analysis does not. Thus, the focus on customer value helps to prevent shortsighted decisions that meet an immediate goal but harm the organization as a whole.

For example, one part of a simulation model will track how many customers interact with customer service and how they behave afterwards. This means that assessment of any proposal to reduce customer service costs will include the impact on later customer purchases. A return on investment calculation can incorporate such factors, but only if the analyst is clever enough to identify them. With a customer value analysis, these factors are included automatically. Similarly, the customer value model will highlight the downstream impact of business plans such as new acquisition programs, avoiding unexpected bottlenecks in distribution or service should volume suddenly increase.

Looked at another way, the customer value approach means the company is using the same framework to assess marketing treatments as staffing levels or business policies. This simplifies the task of the analysts and managers, who can work with a single tool. More importantly, it ensures that all opportunities are evaluated thoroughly and consistently.

The simulation model also provides a comprehensive inventory of customer interactions, which can be used as a checklist of opportunities for business improvement. Having a comprehensive list is more important than it may seem because managers cannot otherwise be certain they are considering all of their options. Today's businesses are so complex that even experienced managers can no longer be confident that they are intuitively focusing on the most important choices, or that they correctly understand how these choices interact. Thus, a proper customer value model provides immense value for corporate planning.

A customer value model supplements, rather than replaces, tactical devices such as predictive analytics. More precisely, it provides a context to ensure that tactical decisions take into account their strategic consequences. Of course, finding the correct strategy itself is still a challenge; the customer value model cannot ensure that a company gives the right answers. However, it does help the company to ask the right questions, and that is an excellent start. 

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