Marketing analysis using RFM (Recency, Frequency, Monetary value) has been used for over 35 years to target the segment of customers most susceptible of answering a marketing stimuli. It is often considered for emarketing segmentation, mostly for email campaigns.
One of the interesting aspect of RFM comes from it's simplicity and ease of application. From historical purchase data, each customer is classified in quintiles (20% slices) by looking at the last purchase date (Recency), the number of purchases (Frequency), and the overall purchase value (Monetary). The Pareto principle applies in most cases: we often find out that 80% of the purchases comes from the 20% of the customers.The "cube" picture to the right, borrowed from Jim Sterne's classic paper on e-metrics, gives a clear representation of the usefulness of the RFM classification.
However, RFM also has it's limitations. Although past behavior might be an indication of future attitude, "tripwires" should be set carefully to avoid frustrating the user. There's also a risk of over soliciting the same cream of the crop over and over again.
A potential customers typically "converts" when a purchase is completed. However, in the web world and depending on the web site objectives, the "conversion" could be any significant business event. On a ad-based web site, the monetary value could become the click-trough rate and the ad-revenue generated. A support web site could evaluate the value of "cost avoidance" of self-service vs. call-center based interactions, etc.The next post will provide a perspective on web site engagement effectiveness.