RFM for email senders – a step-by-step guide to a more strategic messaging approach

RFM stands for Recency, Frequency, Monetary Value. It’s a simple and yet effective customer scoring algorithm that has been used for about a hundred years, now. Especially in catalogue marketing, because RFM saves costs: The method discriminates possible responders from non-responders before the send out. RFM scores correlate with the likelihoods of responding to the next offer.

Email marketers do also use RFM analysis. Less, to save costs (email is “for free”), but more to predict subscriber values and to:

  1. cluster subscribers into homogenous groups
  2. apply targeted messaging strategies to relevant groups
  3. and analyze subscriber migrations between groups

RFM explained

The RFM score for each customer depends on three factors. Those are, in the order of their importance:

  • Recency:
    How long has it been, since the customer bought for the last time? The further back this date lays, the lower the recency score, and the slighter chances of converting after our next advertisement. Out of sight is out of mind. Recency is – in the classical model – the strongest predictor of whether someone responds.
  • Frequency:
    How often did the customer buy in the reporting period (month, quarter, year)? The rarer, the lower his frequency score. A customer, who bought several times, will respond with a higher likelihood than someone, who just bought one or two times.
  • Monetary Value:
    How much did he spend? The smaller the turnover, the lower is the monetary value. Customers, who invested a lot of time and money are worth more – and they respond more likely.

Mail order companies traditionally look at sales values. Alternatively, one could use email response data such as clicks, website visits, or conversions. The RFM scoring model is also customizable. Marketers built numerous variations. Some append a fourth variable like email activity. Others just look at a two-dimensional score concentrating on the two most important behavioral predictors: Recency and frequency.

Calculating scores

A simple algorithm to determine subscriber scores would look like this:

  1. Extract the transactional data for the reporting period from your ecommerce solution, and make sure it contains R, F, and M for each customer:
    1. days since the last purchase
    2. number of purchases
    3. revenue with the customer
  2. Order all values and build groups for R, F, and M. Marketers often choose quartile values as interval boundaries. This leads to four groups of the same size for each of the three variables. For example, recency group one contains the first 25% of the best customers, group two the second 25% and so on. Assign each group a value ranging from 1 (bad) to 4 (best). In the end, every subscriber is assigned one R, one F, and one M value.
  3. Combine your R, F, and M values for each customer. This will result in 64 segments. Segment 4-4-4 contains the most valuable customers: highest recency, highest frequency, and highest monetary values. Down to the inactive 1-1-1 cell, there are fine differences that demand for different messaging strategies.

The result of the procedure looks like this in Excel:

Columns A, C, D, and E hold the export data from your ecommerce solution. R, F, and M to the right hold the formula “=MATCH(C2,QUARTILE(C:C,{0;1;2;3}))” (without quotes, exemplary for the recency column C (“C:C”) and row 2 (“C2”)). Recency has to be negative in order to assign a four to top customers. The last column, RFM, contains the formula “=CONCATENATE(F2,G2,H2)”. This just combines all three scores into one cell.

Relationship management

Let’s have a look again at the above figure: I highlighted two customers. One from the 444 cell (best), and one from the 111 cell (worst). Both are obviously on very different stages of the customer lifecycle. One is highly active, the other is inactive. Thus, both should be treated differently.

By looking at the numbers, we can easily spot some interesting segments:

  • 444 – top customers & target segment for all others:
    Bind those customers and carefully expand the relationship . They are your most valuable assets. Reward their past engagement for example with exclusive offers and information. Try avoiding incentives optimize your margins. Ask for feedback to show customer focus and appreciation for their efforts. Try to connect on other channels, too, like Facebook or Twitter, to intensify the relationship even more.
  • 111 – “flops”:
    Build (re-)activation campaigns and try to get to know, why these are inactive. Perhaps it is just a lack of trust in your company. Or they don’t know about what makes you different? In this case, you could try communicating more clearly things like guarantees, return rights, test and trust seals, references, and so on. Or maybe you just don’t reach them because of deliverability issues? If you can’t activate them be all means, it’s time for a final re-opt-in campaign.
  • 441 – MMPs, many mini purchases:
    Customers in this cell ordered often and recently. However, they generated not much revenue. What would lead to bigger shopping carts? Think for example of coupons that can be redeemed at a certain minimum order value. Perhaps, you could also incentivize them to do purchases, which they have likely planned or considered in the future, already today (cross selling). You don’t necessarily have to buy recommendation engines for doing that. I’ll cover association rule mining in another post.
  • xx4 – spenders:
    These are the ones who bring in the revenues. Try sparing incentives and offering luxury articles
  • 144 – inactive tops:
    They bought often and spent much once. However, their last purchase lies far back. What happened? Build a powerful reactivation campaign for this potentially valuable segment to find out and win them back.
  • 414 – new tops?
    Another valuable segment – they bought some days ago and spent a lot. The cell usually holds customers with a potentially high customer value. Develop the relationships!

Analyze migrations

Of course, it would be nice if we got all customers in 4-4-4 cells: Don’t loose existing ones, bring in all the others. That’s an ideal, our messaging goal. In order to get close, campaign management needs controlling. That means we carefully have to watch the movements of customers between the cells. If customer migrations don’t meet our goals, we have to make adjustments. Therefore, update your numbers regularly.

A Pivot table makes migrations visible in Excel:

Figure: The 444 segment (best customers) grew – nice!


Figure: Some just look at recency and frequency – raise the recency values of low scoring segments and then try to bring them into the direction of 44.

RFM is just one out of many segmentation methods. Plus, there are several cool adaptions. Be the 1st to know about new trending techniques - I'll keep you posted: (archive♞)
Yes, I accept the Privacy Policy
Delivery on Fridays, 5 pm CET. You can always unsubscribe.
It's valuable, I promise. Subscribers rate it >8 out of 10 (!) on average.

3 Responses to RFM for email senders – a step-by-step guide to a more strategic messaging approach

  1. RFM for email senders – a step-by-step guide to a more strategic messaging approach http://t.co/a3K1ihRR via @LukeAnker

  2. RFM for email senders – a step-by-step guide to a more strategic messaging approach http://t.co/6JnbtmlL

  3. Pingback: Meer conversie uit je e-mailmarketing? Zet de RFM-analyse in /  mediafeed.gertimmer.nl

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.