David Raab Article Archive

Building a Customer Value Model
David M. Raab
DM Review
October, 2005

Last month’s column discussed the advantages of customer value models as tools for making business decisions. The basic point was that such models calculate the full impact of a decision by tracing its impact on the customer’s subsequent behavior. Thus a company can avoid decisions that look good in the short run but ultimately decrease the value of the customer relationship.

In other words, a customer value model is a tool for estimating customer lifetime value, which is defined as the net present value of future cash flows associated with a customer. Some readers may be disappointed by this clarification. After all, the concept of customer lifetime value has been around for a long time and received quite a bit of recent attention. If “customer value model” were really the same thing, there wouldn’t be much new to say.

But although customer value models and customer lifetime value are related, they are not identical. Important aspects of customer value models are not covered in typical discussions of customer lifetime value.

Discussions of customer lifetime value tend to focus on how and why to use it. The details of how to calculate it are usually ignored or relegated to an appendix. The few discussions that do take place often focus on shortcuts to provide a reasonable estimate with a minimum of work and data. This is because a key objection to customer lifetime value is the difficulty of calculating it correctly.

By contrast, calculating lifetime value is a primary purpose of a customer value model. How to do it is therefore a central topic of discussion. And although some shortcuts may be appropriate, it’s generally true that the more detailed and precise a customer value model is, the better.

So what does a customer value model actually look like? Simply put, it is a simulation of the interactions between a customer and an enterprise. Like any simulation model–think traditional business process modeling–the customer value model uses a simplified representation of the activities it describes. This distinguishes the customer value model from other methods of estimating lifetime value, which include statistically-derived predictive equations or simple mathematical formulas using a few key variables (typically acquisition cost, profit per year, attrition rate, and discount rate).

Customer value models are built from events. Each event has at least two attributes: a financial value and a relative time. The financial value is used to calculate cash flow. The time value is needed to convert cash flow to net present value and to locate the event in the sequence of the customer’s experience. This sequence is an essential feature of a customer value model, providing insights that a predictive equation or mathematical formula cannot.

In a highly simplified customer value model, one event might combine many separate interactions (offer, response, purchase, product delivery, collection, payment, service, etc.) into a single transaction. A more detailed model would treat each of these interactions as a separate event. Either way, events can be linked to other events. This linkage might be implicit, by simply including both events in the model. The financial value attached to each event would reflect both the profit or cost of the event and the number of customers expected to participate. Or the linkage might be explicit, meaning the events are logically linked, with a time lag and conversion ratio, so the system can itself calculate when and how many customers who participate in the first event will also participate in the second event. For example, the model might link a purchase event to a renewal event, with a fraction indicating that 80% of purchasers will renewal one year later.

Event linkage is a critical feature of customer value models. It is what allows the model to trace the ramifications of an event through the balance of the customer relationship. Isolating the impact of an event is typically done by running a scenarios of the model with and without the event in place, and then comparing the differences in the results. This resembles conventional test vs. control studies, except that it relies on simulated rather than actual results.

Because the customer value model is a simulation, users must bear in mind the distinction between precision and accuracy. Results of a customer value model are quite precise, meaning that small changes produce consistent, measurable differences. They may or may not also be accurate, in the sense of matching the real-world results. In fact, because the model is predicting future events which are subject to many external influences, complete accuracy is impossible. But the model is still useful so long as users accept its outputs as meaningful. In particular, they must accept that the differences between scenario results accurately reflect the impact of the changes in the scenarios themselves.

More detailed models are more work to build, but they also allow more precise simulations. Thus, assuming actual data or credible estimates are available, a detailed model can calculate the effects of slight changes in customer treatments. This makes the customer value model a tool for tactical decision-making, in addition to broader strategic analysis.

The customer value model has other tactical uses as well. The value attribute can be broken into revenue and cost components for detailed financial analysis. Current-period quantities can be assigned to each event at the start of the model, thereby producing forecasts of future quantities by period. Additional attributes can forecast quantities such as inventory requirements, call center volume, and renewal contacts. Operational managers can evaluate the impacts that changes within their departments have on the rest of the company. These impacts can be financial (reducing customer service might save money but result in lower future sales) and operational (more accurate order entry could reduce the number of returns processed at the warehouse). This is important because decisions that make sense for one department often have negative impacts elsewhere. Without a customer value model, it is difficult for department managers to see the full picture.

Customer value models can also serve a still larger role as marketing planning and management tools. Models inherently create an inventory of the interactions between a company and its customers, so they can provide a central platform to review, document and modify business rules associated with those interactions. The time element of the model means it also provides a framework to organize the list of planned promotions with their expected results. If the model is used for forecasting, some of the promotion plans must be captured anyway. Similarly, actual results, which are posted to a model to check its accuracy, can also be used to generate marketing performance reports.

These more advanced functions are not necessarily present in today’s customer value modeling software. In fact, much customer value modeling is done on ad hoc spreadsheets. Sophisticated customer value modeling systems have long been available for specialized applications including magazine circulation and catalog modeling. Some modeling features are available in marketing management systems and optimization software. Otherwise, customer value models must be built using general purpose business process modeling and simulation systems.

This can be expected to change as the increasing interest in customer lifetime value leads companies to demand more precise calculations and increases their willingness to pay for them. Marketing management software vendors will expand their modeling capabilities and process modeling vendors will add specialized customer value modeling features. Dedicated customer value modeling systems will likely appear as well. This is good news for companies in search of solutions, although it will be some time before standard solutions appear. Interesting times are ahead.