Tuesday, September 19, 2006

The New York Times Discovers That Marketers Use Science

Today’s New York Times carried an article on the use of science by marketers, although this is not exactly news. (“Enlisting Science’s Lessons to Entice More Shoppers to Spend More”, Science Times, September 19, 2006.) The specific examples it will be familiar to any marketing professional: carefully tracking shoppers’ paths through a store; analyzing physiological reactions to advertisements; and using multivariate test designs to assess ad components.

This last example brought to mind what I consider a more interesting (and somewhat contradictory) trend: a growing recognition that it’s no longer viable to attribute customer actions to a single promotion. That was the traditional approach taken by direct marketers: put a code on the order coupon and measure results by counting the returns. We always knew this was an oversimplification, since customers receive multiple promotions and the coupon they happen to send back isn’t the only one that contributed to their action. But direct and database marketers have always been so proud of being “measurable” that it would have been hard to admit their primary measurement technique was more than a little dicey.

Still, the proliferation of channels and message opportunities makes it impossible to continue ignoring the contribution of all those other contacts. This means we need something similar to multivariate test design to help manage the customer experience. That is, we must assemble information on all the messages each customer has (or might have) been exposed to, and then analyze how the different messages correlate with different results. In the past, direct marketers did this in a very controlled fashion by creating test panels that received different streams of messages. That’s no longer practical in a world where there are so many potential combinations of message frequency, channel and content, and where message delivery is often beyond the marketers’ control. Instead, we’ll need to rely on techniques like multivariate testing to come up with estimates that may be crude, but are still more realistic than pretending multiple messages don’t exist.

Incidentally, this is a partial answer to one of the questions I posed yesterday. Multivariate analysis is one of the ways we can understand enough about complex customer behaviors to build an effective customer experience model.

1 comment:

Jeremy Els said...

The Internet presents an unprecendented opportunity for large scale message and promotion testing. There are, in fact, at least several commercial products that provide this capability, e.g. SiteSpect (www.sitespect.com). Combined with sophisticated web analysis tools, Internet marketers indeed have alot of scientific tools at their disposal.