|
RFM WAS, IS AND MIGHT CONTINUE TO BE THE STANDARD FOR CUSTOMER SEGMENTATION in the direct marketing industry. The logic behind R (Recency) and F (Frequency) is solid -- but can M (Monetary) be improved upon? Probably! |
When using RFM to segment customers, usually the most frequent customers are best and the highest monetary customers are best. However, high Monetary segments do not always outperform low Monetary segments. These notable exceptions tend to occur when the Average Order is substantially different between segments. Quite often -- perhaps most often -- the segment with the highest Average Order outperforms the segment with the highest Monetary. Following are a few case studies that pit RFM against RFA. The results may surprise you.
IS AVERAGE ORDER A BETTER PREDICTOR THAN MONETARY?
With any customer segmentation technique, it is assumed that past is prologue. The number of times a customer bought in the past is important. So is the amount of money the customer spent on each purchase. Many buyers exhibit a comfort level of how much they will spend with each purchase. For some this may be $50 and for others it may be $100. The fact that a customer spent $50 on each past purchase is a powerful indicator that he or she will continue to spend $50 on each future purchase.
The relative importance of each variable, whether it is R, F, M or A, varies from one customer base to another. In the following examples, Average Order proves to be more valuable than Monetary in segmenting customers. Could any of these match your situation?
FREQUENT GIFT BUYERS WERE NOT THE BEST!
A gift cataloger was segmenting using RFM. When the segments were created and tracked, the veteran marketers who devised the communication plan were surprised by the results. Buyers with a Frequency of 1 and a Monetary of $50 to $100 outperformed buyers with a Frequency of 2 and the same Monetary. This happened regardless of the Recency segment.
"Old School" logic dictates that, given the same Recency and Monetary, the higher Frequency segments are best. In this case, however, the opposite was true. In fact, the Average Order was only half as much for two-time buyers as it was for one-time buyers. The difference in response rate was small, but the segments with the larger past Average Order continue to have larger Average Orders. As a result, compare to two-time buyers, sales per contact were much higher for one-time buyers with the same Monetary.
FINDING UNPROFITABLE "BEST CUSTOMERS" WITH RFA
A collectible cataloger planned to use RFM to identify its best customers and offer them a special "thank you". The cataloger started with standard logic: Make a list of all buyers who spent several hundred dollars, bought in the last six months and made three or more purchases. While the cataloger created the list ob buyers, it printed the actual RFM (date, number of purchases and dollars) next to each name and address for management review. The cataloger was shocked by the results!
A large number of "Best Customers" actually made many purchases (15, 20 or more) and spent only $10 or $20 each time. Given the cost of fulfillment, the cataloger actually lost money on each sale. Instead of being "Best Customers", they were "Worst-Nightmare Customers" -- at least from a profitability standpoint. Needless to say, the cataloger developed a new criterion using Average Order as a parameter for creating new offers.
MAKING BETTER OFFERS WITH RFA
One of the best examples of the power of RFA I have encountered is from a retail jeweler who segmented customers using Average Order. The jeweler made the same set of offers to a portion of each Average Order segment and measured the results. The jeweler surmised that buyers were likely to spend predictable amounts of money each visit. Buyers of inexpensive jewelry7 ($50 average sale), for example, behaved differently than buyers of "expensive" jewelry ($500 and more per sale.)
The jeweler separated the $50 average-sale buyers from the $200 average sale buyers from the $500+ average-sale buyers. Then, customers in each segment received one of three offers. A control group got nothing. Some got an offer for a percentage off orders of $50 or more and some got an offer for a percentage off orders of $500 or more. Here is what happened.
The $50 buyers responded best to the $50 offer. Those $50 buyers who received the $500 offer actually performed worse than the control group that got no offer. The $500 buyers responded best to the $500 offer. Those $500 buyers who received the $50 offer performed worse than the control group.
Because this jeweler was blessed with a large number of frequent buyers, RFA helped the jeweler design offers that fit its customers' spending habits better than RFM. Had they used RFM only, they would likely be making higher-dollar offers to frequent but low Average Order buyers who were nevertheless profitable.
SEGMENTING LOW AVERAGE ORDER SEGMENTS BETTER WITH RFA
This example comes from a catalog/retailer that uses space ads in its
media mix to attract new buyers.
The company was treating low Average Order buyers from its space
ads in the same way it treated low Average Order buyers from its
catalogs. The problem was that a below-average catalog Monetary was
less than $50 and space-ad orders were around $20. A space-ad buyer could
place two or three orders and still have a total Monetary below
the average catalog order.
For this company, examining Average Order presented a finer distinction in low-dollar buyers as well as a clear split between space-ad and catalog buyers. After the distinction was made, it was clear that space-ad buyers were returning with orders that were well below average and their Lifetime Value was much less than that of new buyers from the catalog. As a result, the company made different offers to low Average Order buyers to encourage them to spend more each time.
COMPARING HIGH FREQUENCY BUYERS WITH RFA
In business-to-business applications, the simple Frequency segmentation of one-time, two-time and three-time+ buyers often does not work. This is because business buyers may buy many times from each supplier. Although more orders are good, many small orders can be costly.
In a review of one company's customers, the company noted that its top client spent $12 Million in fiscal 1999 and it's No. 2 client spent just more than $10 Million in fiscal 1999. Based upon RFM, both looked great.
With a little more probing, it was noticed that the top buyer had a much lower Average Order than the No. 2 buyer did. In fact, the top buyer placed 159,000 more orders in one year than the No 2 buyer! Now, $2 Million is a lot of sales, but handling 159,000 orders is a lot of fulfillment cost. The managers of the business-to-business seller had no idea that the two companies were behaving differently. Overall profitability is quite different from what RFM might suggest.
Nonprofits have used RFA for a long time. Would it be reasonable for a charity to expect donors who have a Recency of one year, a Frequency of five times and a Monetary of $100 to give $100 each next year? Experience tells nonprofits to ask for $20, or in some cases, $25.
Charities have known for a long time to ask for an amount their constituents feel comfortable with for each gift. Doesn't it make sense that for-profits would do the same for each sale?
WHICH WILL WORK BEST FOR YOU?
The best way to determine if RFA works better for you than RFM is to test both head-to-head. To determine if a test is worthwhile, review your database to answer these questions:
- Is there a large difference in Average Order among "Best Customers"?
- Do most of your buyers have a low Monetary and a Frequency greater than 1?
- Do your customers tend to order a consistent amount? That is, high-ordering customers stay that way?
- Do more than 1/3 of your customer have a Frequency of 1 and a Monetary below the Average Order?
If you answer yes to one more more of these questions, it makes sense to test RFA. If you need help identifying segments or making offers, find someone with the experience to assist you. You may be in for a real surprise -- and you might just drop RFM for RFA!
Questions? Just ask! [Back to How-to Resources] [Top] |