Marketing Profs

You may or may not be using the basic segmentation strategy of RFM (recency, frequency, monetary)—that is, dividing your mailing list into a few buckets based on recency in ordering or visitation to the site, the number of times they've ordered or visited the site, and the lifetime spend.

My issue with RFM models is that I would instead like to see each threshold between activity, and tweak it on an ongoing basis. That's the joy of email marketing, it's all so available and adjustable, and in real time.

The common issue I see is that it's difficult to report against the nice divisible cells we've made; so, before we continue, first make sure that your ESP can report back to you performance by cell.

Ideally, your ISP (or you, if you are in-house) can do response filtering: e.g., "Did not click on offer last week, was in Cell X." (Responsys has separate filters from segments, and YesMail has cell reporting if you ask for it.)

Analysis

Make sure you do some analysis on your customer base before breaking down the groups. Creating a group of revenue buckets divisible by 10 makes nice round numbers, but do they really relate to your customer's spending behavior?

What is the goal? Is it more product variety, more new customers, higher spend, more average spend, more lifetime spend...?

Let the data tell you (instead of you telling the data) where those boundaries and thresholds are, between each stage.

Examples

Let's say that company A is an RSS feeder. It notices that if customers set up three feeds they tend to stay with the customer far longer than those who set up two. The segmentation strategy would therefore separate along that line: those who set up two, and those that set up three.