Moving beyond RFM analysis to create truly data-driven customer segments
RFM stands for recency, frequency, monetary value and is a method for calculating or assessing overall customer value. Its most common use is in direct marketing or database marketing and it is particularly heavily used by retailers and charities. RFM analysis is predicated on the idea that your most valuable customers are those who spent most recently, who spend most often (frequency) and who spend the most (monetary value).
Most organisations hold this kind of data about their customers which makes RFM analysis relatively accessible and easy to do. Typically each RFM measure gets a score between 1 and 10. These scores can then be added together or represented in other numerical ways to give each customer a total RFM score. From these scores segments can be created and collapsed together, meaning that each customer has their own individual score as well as a score denoting membership of a segment containing other similarly-scored individuals. A segmentation model is born.
Advocates of this technique point out that it has the virtue of simplicity: no specialized statistical software is required (although such software is inexpensive and can dramatically simplify the process) and the results are readily understood by business people. In the absence of other targeting techniques, it can provide a lift in response rates and has done for years.
However, RFM analysis has a number of substantial limitations. Firstly, it is purely descriptive so does not provide a mechanism to forecast things like next best offers in the way that a predictive model might. Secondly, an RFM model assumes that customers are likely to continue behaving in the same manner – it does not take into account the impact of life stage or life cycle transitions on likelihood of response. Thirdly, if an RFM model is the only way of differentiating between segments then this often leads to over-marketing to the most attractive segments and neglect of other seemingly less attractive segments which could actually be profitable if they were properly developed. Finally, an RFM model is based on pre-determined business rules – you decide what counts as recent or high-spending, often on a fairly arbitrary basis, and the model sorts customers accordingly rather than telling you which patterns of behaviour are significant.
So, how can you move beyond simple RFM segmentation? The next step is to use true clustering techniques to develop data driven segments of customers. Rather than a pre-determined business rule approach such as RFM, a true clustering approach can take all available customer data (including demographics, attitudes, transaction patterns, specific product groups, payment history, channel patterns and so on – whatever data you have can be used) and use all of it to create a set of meaningful clusters based on the variables that really predict behaviour rather than on arbitrary business rules. This type of modelling is much more powerful than RFM analysis.
Using this approach organisations can build different segmentations for different campaigns or different aspects of the business. For example, you might want to segment customers according to which groups are most likely to respond to a particular promotion, and then re-segment to uncover groups that are most likely to buy a new product. These segments might be very similar, but they might not be and the differences between them would not be picked up by a simple RFM analysis. A cluster analysis can have ‘something in mind’ while creating its groups and another version of techniques in the predictive analytics toolbox can help to optimise which promotions are sent to which people in order to combat the “best segment” problem.
RFM has been about a long time and if you’re not doing any analytical targeting at all then it’s a really great place to start to get quick wins and prove that the approach works. If you have have been doing it for a while, however, it’s probably time to consider branching out and using the rest of your resources to truly understand what your customers are telling you about their segment membership. Smart Vision can help you learn more about what’s possible using the data you already have.