Marketing is a complex process, and getting more so. As technology permits us to track more customer data and to interact more cost-efficiently with individual customers, the choices open to marketers has skyrocketed. Tesco, the UK supermarket chain, began a frequent shopper program more than ten years ago, and today they have fairly good records on the grocery shopping habits and preferences of some 11 million customers. Their quarterly newsletter goes out in 5 million different versions, which is about as close to individualization as anyone can get. The quarterly mailing generates a response rate among members in the 25% to 30% range, which means there is an awful lot of additional information created with each campaign.
Evaluating this kind of interaction data has also become more complex and difficult. In the world of direct mail campaigns, the way you decide what offer to make and whom to make it to is by testing. Send out 10,000 mailings of Offer A to one group of folks, and 10,000 of Offer B to a different, but statistically identical, group. Then wait for the responses to decide which offer is best, and by how much. But when you have, say, 2 million registered users, each with an individually specified profile, checking your Web site every week or two – how do you use testing to learn which offer makes the most sense for Customer 3675-J?
As if this new dimension of customer-specific marketing isn’t complicated enough, just running a traditional advertising program is more complicated than ever before, as well. I talked to one ad manager at a big-brand US retailer who told me that four or five years ago his company would run 30 to 50 campaigns in a year and they thought they were doing very well. Last year, however, they ran over 2,000 campaigns, and he expects to be running several thousands of “campaigns” aimed at mini-segments and smaller and smaller markets in just a year or two.
Add to that the increasing sophistication of statistical analysis that has become possible with information technology, and traditional research data takes on a new complexity all by itself. Graham Hill’s comment elsewhere in this blog (http://unicashare.typepad.com/share/2006/09/is_customer_ser.html#comments), for instance, suggests that in evaluating customer attitudinal data such as satisfaction, researchers must now use statistical tools designed to track measurable levels of interaction among the data – because a single customer’s attitude changes through time based on his previous attitude, and also because one customer’s attitude can sometimes influence another customer’s attitude. Call it what you want, but as Graham comments, “Life is complicated.”
The obvious question here is how do marketers keep all this straight. How, in fact, do we marketers assess results and prioritize efforts? I’d be very interested in hearing from people how they deal with this increasing level of complexity…