446 CHAPTER 13 DIRECT MARKETINGSome publications have criticised the RFM score model as a poor way to measure loyalty.
One problem is that patterns of buying behaviour for frequently purchased goods are different
than those for infrequently bought goods. RFM ignores the average time between purchases
as a key variable, while the probability that someone who is within historic range of buying
frequency will buy again in the future is higher than for someone who is way past the average
time between two purchases. RFM analysis would determine that someone who buys more
frequently and has bought recently is more loyal, and therefore direct marketing activities
and investments would be targeted at the wrong profile of customers. Take, for instance, two
customers, Mr Smith and Ms Jones, who both start to buy goods from a company in month 1.
During the first year they purchase at different rates: Smith buys again in the second, sixth
and eighth months, whereas Jones purchases again in the eighth month. A simple RFM
analysis suggests that Smith is more loyal and thus more interesting for direct marketing invest-
ments than Jones because his purchases are more frequent and recent. But Smith usually buys
every 2.3 months and yet by month 12 he has not bought anything for four months. Jones,
too, has not bought anything since month 8, but she normally does not purchase anything for
seven months. On this basis, the chance that she will buy again in the future is higher than
for Smith. This is a case of event-history modelling. In its simplest form, the formula to calculate
the probability that a customer will keep on buying is tn, with n the number of purchases that
the customer made during a period (in this case a year) and t the fraction of the period rep-
resented by the time between the customers’ first and last purchase. Unlike RFM, this model
is particularly good at predicting how soon a customer’s buying activity will drop off and
might prevent heavy over investment in profitable but disloyal customers.
The second main disadvantage of a scoring model such as RFM is that the monetary value
is mostly based on revenue rather than profitability. By multiplying the probability figure
for each period (e.g. a quarter) by the historical average profit number, the sum will be the
estimated profit for each customer over the next year.
After analysing the customers’ profitability and the projected duration of the customer
relationship, all customers can be placed into one of the four categories in the matrix shown
in Figure 13.7.Figure 13.7 Reinartz and Kumar’s matrix for categorising customers and relationships
Source: Reprinted by permission of Harvard Business Review. From ‘The Mismanagement of Customer Loyalty’
by Reinartz, W. and Kumar, V., July, 2002, pp. 86–94. Copyright © 2002 by the Harvard Business School Publishing
Corporation; all rights reserved.M13_PELS3221_05_SE_C13.indd 446 6/5/13 2:56 PM