DM News Essential Guide to Lists and Databases: Make Your Database Marketing More Scientific
There are three general methods for more effective and relevant database marketing. First, predictive modeling can save money by creating more targeted communications. Second, optimizing can help an organization do more with less. Third, more effective test marketing can make communications more efficient.
Predictive modeling. Building predictive models is one of the best ways to become more targeted. The process of creating models need not be intimidating, as the concept is simple: Use historical data to predict future behavior. In database marketing, this often means looking at the variables of those customers who responded previously and determining which combination of those variables is significant.
Once created, a response model that predicts a response rate can be used to determine whom should be contacted. For example, a database marketer might take only the top 30 percent of those customers who are most likely to respond. By knowing that the bottom 70 percent is unlikely to respond, the cost of mailing to them can be spared.
Optimization. Optimizing campaigns is based on the principle of limited resources. We don't have unlimited budget or channel capacities, nor do we have unlimited customers. Given that we can't contact every customer with every offer (and probably shouldn't), we need a mathematical way to determine who should get which offer.
To help make this decision, an optimization algorithm may take as input such things as the probability scores mentioned above, cost per communication, contact policy and budget and resource constraints. The result is an economically optimal offer. As a byproduct, opportunity costs also can be determined, letting the database marketer perform what-if analyses such as, "What will happen to my profitability if I increase the budget for the fall campaign?"
Testing. Traditional testing in database marketing is usually a somewhat simplistic test-and-control approach. The concept behind test-and-control is that it is necessary to have a holdout sample to determine the true effect of the marketing since we know that some customers buy products even when those products aren't marketed.
A new, more scientific approach is emerging in which we don't just test one factor against another, but test many factors simultaneously. This new approach, which uses experimental design techniques, lets database marketers create powerful tests that consider such factors as the color or size of the envelope, in addition to the price point and the creative.
Traditional methods such as champion/challenger let you test only one factor at a time, ignoring the possibility that a combination of factors may have much better results.