Shop Talk:

In a perfect world, you'd analyze this behavior at a customer segment level. You'd measure all demand, then costs (catalog plus search) including all paid search clicks that didn't res-
ult in purchases. If you can do this, the results are fascinating. If you can't, make 
an assumption.

If you have 10 clicks that result in search purchases across a segment of 1,000 customers, you probably have 250 clicks that didn't, so you simply make an 
assumption of the true cost of search that's "caused" by a catalog mailing.