Who Needs Print, You Say?

Of course, matchback results are most accurate if you do a great job of tracking other order sources, such as e-mail and search, so those orders don’t end up in the “unknown code” bucket.

4. Conduct a hold-back test.
Do you suspect that matchback is over-attributing sales to your catalog? OK, try a hold-back test. Split a group of names into a control group that receives catalogs using your usual mailing pattern and a test group that is “held back” for the same period; that is, you don’t mail any catalogs at all to the test group. Keep that test group as small as you can to minimize sales loss, but just large enough to get statistically significant results. You’re likely to see this: