Share this article

Omnichannel commerce | Database Marketing

RFM Modeling: Simple, Cheap, and Effective

image: Your Commerce
This database marketing technique lets you maximise your customer database without making a major investment

Everyone in business hears a lot about data these days.

 

We hear how much of it there is. We hear how important it is. We hear how many companies provide software to hold data, move data, process data, report data, analyse data, and connect data to more data. Big data can be immensely useful and profitable... for those with big budgets.

 

What about the rest of us?

 

For the marketer without a Ph.D. in data science or a large budget to spend on tools and services, what can be done? I have always told my clients to start in one critical place: where the money is. Who are your best customers, and how can you make them even better?

 

For direct marketers, going back decades to pre-ecommerce catalogue merchants, there's a simple exercise that can do wonders: RFM modeling. The initials R, F, and M stand for recency, frequency, and monetary value, the most valuable data about your customers. By segmenting customers using these three data fields, you can identify your best (and worst) customers and formulate a marketing strategy.

 

As a start, I recommend assigning customers in your database to one of three levels—high, medium, and low—within each of these three data fields. Keep it simple and just divide your customers into thirds. So in your database, record which customers are the third who have purchased more recently, the third who purchased next most recently, and the third who purchased least recently. Like any recipe, modify to taste; for instance, you could just assign customers to the levels "past 30 days," "30-90 days," and "over 90 days" instead.

 

For frequency, again assign high, medium, and low based on the number of times each customer has purchased (perhaps limiting the time frame to the past year or so). And for monetary value, assign based on the total sales amount over that same time period. Given that you have three levels of three data fields, you now have nine customer segments, each consisting of just over a tenth of your total customers.

 

These segments are a starting point for a marketing strategy. For example, segment MHH (medium-level recency, high-level frequency, and high-level monetary value) is a segment of customers who have been frequent purchasers and spent a lot, but haven't purchased in a while. They may be at risk of switching to a competitor, so a strong offer to lure them back could be worthwhile.

 

What about segment HHL, those who are high in recency and frequency but low in terms of monetary value? You should look at these customers carefully. Are they low spenders who actually cost you money in customer service? Depending on your business, they may be customers you are happy for your competitors to take, or maybe you can find some hidden potential and turn their frequent shopping into a stronger revenue stream.

 

The segmentation does double-duty as a targeting model. By ordering the segments from HHH down to LLL, you have ordered your customers in their likelihood to respond to a marketing campaign. While an e-newsletter may not cost much, if you plan on, for example, sending a postcard to your customers, the vast majority of responses will come from the top of this ordered list. By mailing only the top segments, you reap almost the same number of responses while substantially cutting the cost.

 

For example, assume that the uptake on a special deal offered via postcard follows the 80/20 rule (if you don't know the 80/20 rule, look it up!), with 80% of the responses coming from the top 20% of the customers. In terms of the ROI on our hypothetical postcard campaign, we would be multiplying our return (the numerator of our ROI) by 0.8 (because we do miss a few of the potential responders lower down the list). But we multiply our investment (the cost, and in this case the denominator) by 0.2, because we decided against sending lots of unnecessary mail). In other words, we multiplied our ROI by a factor of 4.

 

For the highly sophisticated, well-funded data-driven marketer, the opportunities to benefit from data are nearly endless. But a simple and straightforward approach to understanding customers and their value is the foundation of any marketing analytics program. Identify best customers, and treat them well. Find customers with potential, and nurture the relationship. Recognise customers who cost more than they are worth, and let your competitors have them!

 

Start with the time-tested RFM approach and take time to become comfortable working with it. When you've gotten as much out of RFM as you can, then you can worry about taking the next step and going further into the world of data and analysis, but there's no rush. Good execution based on simple analysis is much better than fumbling efforts emerging from overly complicated analysis.

 

The goal is never cool algorithms and fancy data analysis; the goal is bottom-line results. (But I must admit, it is fun when cool algorithms get you there.)

author: Michael Bagalman

Michael Bagalman

Michael Bagalman is CEO of Paradox Resolution, a boutique consultancy working at the intersection of data science and marketing.

 

Share this article