Triggers: Weighting Transactions

From time to time, I work on projects where the goal is to set up various triggers.  This can be a complicated problem, because customers might perform competing activities.

I run regressions against various purchase, visitation, and social activities.  My goal is to identify when an activity should trigger a marketing tactic.  You learn some interesting things when you do this (your mileage will vary):

1. Old-School Catalog orders have a half-life of maybe 24 months.
2. E-Commerce orders have a half-life of maybe 20 months.
3. In-Store retail purchases have a half-life of maybe 16 months.
4. A click through an email campaign may have a half-life of 2 months.
5. A visit to a website may have a half-life of 2 weeks.
6. A social media action may have a half-life of 2 days.

Again, your mileage will vary.  But it's your job to know the half-life of all activities!  If a customer visits your website on December 13, but purchased in a retail store on December 1, it is quite likely that the retail transaction will carry more weight ... any triggers you plan are focused more on the retail transaction than the website visit.

Know the half-life of all customer activities.

More On Modified RFM For E-Mail Targeting

Many wonder what to do when you have two potential versions of an e-mail campaign. Which version should a customer receive?

Companies loaded with analytical talent have interesting algorithms to make these decisions. Yesterday, we talked about a shortcut that gets us 80% of the benefit for about 5% of the work.

But what do you do when one version of an e-mail campaign is so much more productive than another? In other words, say you have a Mens and Womens version of an e-mail campaign, and the customer could receive either version, but the Mens version is much less productive (sales per e-mail) than the Womens version?

A shortcut is to evaluate the historical difference in productivity, and apply that to the "weighting" score from yesterday's post. In other words, if the Mens version performs at 65% the level of a Womens version, multiply your Mens weighting scheme by 65%.

Again, this is statistical blasphemy. But you don't work at a company where you have thirteen statisticians sitting around waiting for new and exciting challenges. You're lucky to have one good analyst, and the demands upon this person's time are many. So take the shortcut, and get 80% of the benefit for 5% of the work. And when you have the money and/or human resources to do e-mail targeting the right way, by all means, pursue the ideal answer.