True Marketing Databases Make Sophisticated Data Mining Possible

(This topic was first covered in the June 18, 2010 Direct Listline article, “How Marketing Databases Differ from Operational Databases.” http://directmag.com/lists/0622-lists-how/)
By Jim Wheaton
There is a big difference between a marketing database and an operational database. A marketing database supports sophisticated data mining and an operational database does not.

Sophisticated data mining, in turn, is impossible without the ability to recreate multiple past-point-in-time (“time 0”) views. This is because data mining professionals work in the present, on the past, in anticipation of the future. For example, multiple customer and house non-buyer “time 0” views make it possible to:

*Create the analysis and validation files required for statistics-based predictive models.

*Generate the data for all cohort analysis, including lifetime value.

*Monitor changes in customer inventories, such as fluctuations in segment sizes over time.

Multiple “time 0” views also support data mining to understand how lifecycle changes affect consumer purchase behavior. Direct marketers are lucky because, as a natural consequence of running their businesses, they receive all of the detailed order, item and promotion history required to perform lifecycle analysis. Retailers are not so lucky, unless they have a mechanism for identifying customers and tracking their behavior. That is where loyalty programs come into play.

Let’s take a vertical industry – publishing – and work through a hypothetical example. Keep in mind that, although the specifics are peculiar to publishing, the general concepts are universal across vertical industries. We’ll begin with two assumptions:

*A publisher of a magazine that is targeted to people in their 20’s and 30’s wants to understand how changes in lifecycle affect renewal rates.

*The publisher hypothesizes that renewal rates are adversely affected as subscribers begin to raise families.

If the publisher’s hypothesis is true, then we would expect to see a drop in renewal rates as subscribers move from multiple family dwelling units (MFDUs) to single family dwelling units (SFDUs), or from urban to suburban locations. With a properly constructed marketing database, multiple subscriber cohorts can be analyzed over time for such relationships; that is, from when they first signed up for the magazine though all of their subsequent renewal cycles.

People in their 20’s and 30’s are notoriously mobile. For example, from the time I entered the workforce in 1980 to when I purchased my first SFDU home in 1988, I lived in five different apartments in three different cities and states. Without being able to recreate “time 0” information, it would be impossible to track this sort of customer movement.

The inability to track customer movement is the unfortunate outcome of any marketing database designed such that, every time an address change is received, the previous address is over-written. Such a database will never be able to support data mining to understand how lifecycle changes affect customer purchase behavior, no matter how many years of history have been accumulated.

Does your marketing database overwrite address information as notifications of customer relocations are received? Are you even certain that you have a marketing database? Many companies think they have a marketing database when, in fact, what they really have is an operational database.

Jim Wheaton ([email protected]) is a principal at Wheaton Group.