Last post I showed a graph of our average direct mail response rates over the last few years. I discussed how we classify and segment our customers and how that segmentation has resulted in some phenomenal response rates.
One of the cornerstones of our segmentation model is weighted transactions. The function of a weighted transaction is to make sure that more valuable transactions score higher than less valuable transactions. A direct-mail redemption, for example, is much more valuable to us on average than a sweepstakes signup. The reason for this is that a direct-mail redemption almost always results in a purchase, while a sweepstakes signup rarely does.
Until recently we thought of a direct-mail redemption as a single transaction type. We were so focused on customer segmentation and the benefits it was bringing to our redemption rates that we didn't think to segment our campaigns. Let's revisit the graph from the last post.
This line hides a secret. Look at what happens when it finally occurred to us to break out the response rates by campaign.
Here is a completely different story. Back-to-school and spring redemption rates are through the roof, while our holiday redemption rates are clearly struggling. This graph communicated two things right off the bat. First, we couldn't continue to think of the direct-mail redemption as a single transaction type. When pulling data for a spring mailer, for instance, a holiday redemption is clearly worth much less than a previous spring redemption. So we lost the single transaction type and replaced it with one for each campaign. The second thing this graph communicated was that we needed to rethink our holiday data.
The insight from this second graph allowed us to score customers in our most recent holiday data file using a heavily weighted holiday direct-mail redemption and to assign much less significance to spring and back-to-school redemptions. The result of this is a 2009 holiday redemption rate that is already over 10% and climbing, double what it was last year and respectably above its all-time high in 2007.
What other segments are hiding in my customer data?