|
LOSING CUSTOMERS IS A CONTINUING THREAT to the health of any company for two reasons. First, losing good customers hits profits hard because most profits come from repeat sales to established good customers. Second, it drives up acquisition costs because requires more prospecting for new customers and it costs more to acquire new customers than it costs to keep old ones. |
Nevertheless, some degree of customer attrition is normal for every business. A good marketing plan recognizes this and adjusts for it. Developing a good marketing plan is much easier if you develop a "map" of your company's unique attrition profile to help guide your marketing decisions. Here are some insights about how to create and use an Attrition Model to better manage customer loss.
WHAT IS AN ATTRITION MODEL?
An Attrition Model is simply a specialized kind of predictive model. (Learn more about Predictive Models in general in Prediction - Finding Gold in the Data.) An Attrition Model It provides a way of getting a "big picture" of how your customers behave -- in particular, how the customer segments that you are losing behave.
The most simple Attrition Model is to track Recency -- that is, how long ago each customer made a purchase. You can then apply a the simple rule that the longer the lapse since a customers last purchase, the greater the probability that they will never buy again. (You can learn about many more such "rules of thumb" from Segmentation Secrets and Trade Secrets.)
You can expand the effectiveness of this simplistic model by adding additional dimensions such as Monetary (total spending), Average Order and Purchase Type for each customer segment. The most elaborate Attrition Models also add demographic and psychographic dimensions.
HOW DO YOU USE AN ATTRITION MODEL?
Many marketers regularly recontact all customers who have made a purchase within a certain time period -- typically two or three years. The period may vary based upon seasonal demand, but customers who haven't made a purchase after a certain time are dropped and assumed to be lost through attrition. This is a very simple rule and it is easy to implement. Unfortunately, it is not very cost-efficient.
Consider that all customers are not equal. Dropping all customers who have not bought within a certain time period assumes that all customers are equal -- which is rarely the case. The diagram below shows the profitability of such efforts using customer segments based upon recency and the amount of their first purchases. The dashed line shows the breakeven point below which the organization loses money when recontacting customers.
An example of Segment #4 would be customers with small initial orders. The graph above shows that these customers are unprofitable even if contacted within just a few months. For segments like these, it is likely you will spend more money re-contacting them during the 24 month period than you will make on the initial sales. In other words, the smart marketer will simply let customers in segments like these go on their way. They cost more to keep than they will provide in profits.
For other customers in segments with a higher initial sale (Segments #1, #2 and #3), their higher profits more than make up for the losses in the lower-performing segments. It is foolish to recontact them no more frequently than low-performing customers. There is no reason to stop contacting them at the same time as low-performing customer segments. Smart marketers will make extra efforts to retain these customers and build upon their stronger purchasing behaviors.
DEALING WITH COMPLEXITY
It is easy to graph and understand a segmentation strategy based upon two variables. The real world is much more complex, however. For more than the simplest questions, a marketer needs to use advanced statistical techniques to analyze the multi-dimensional relationships that affect attrition.
For example, simultaneously looking at customer segments upon the basis of Recency, Frequency, Average Order, Size, Product Category, Customer Age and Customer Sex is impossible with a two-color graph on a flat sheet of paper. Nonetheless, all of these factors often affect attrition. A multi-dimensional analysis using these factors is needed to design an effective marketing strategy. The trick is to build a model with segments that "go together" in ways that allow you to treat them appropriately. A good statistical analysis does this and makes them manageable. Otherwise, multi-dimensional analyses are too complex to be of much use.
Grouping techniques, such as CHAID or C&RT are helpful for developing segment definitions that are predictive. These tools split customers into groups that (1) are different enough to act different and (2) are defined well enough to guide decisions about offers and creative messages. It helps if the number of segments is kept to a reasonable and actionable total,. Ten to thirty segments is usually about right, for example. Statistically, there is no overlap between segments but it often helps to combine certain similar segments for marketing purposes. For example, if your Attrition Model has three or four age 18-to-35 segments, one might combine them to get one offer while customer segments age 35 and older might get another.
SEGMENTATION LEADS TO CONTACT STRATEGY
Marketers often begin an Attrition Analysis by looking for areas of waste. Attrition models are good at estimating when customer segments will be come unprofitable to recontact. However, they cannot predict how well a new offer will perform.
Applying the results of a statistical Attrition Analysis will probably lead you to different contact strategies than you have used before. They may be more complex. You may find that you need to make fewer total contacts, but for those you do contact, you may need to make more frequent and more tailored offers. These may easily require more creative and production work than simpler approaches but they will generate higher response rates.
Fortunately, companies tend to underestimate the amount of money they spend on recontacting unprofitable customers. So, overall costs often stay about the same because the additional amounts spent to contact high-performing segments are offset by savings from not contacting under-performing segments
ATTRITION IS NOT JUST A MARGINAL CUSTOMER PROBLEM
Attrition Analysis is most valuable when it is first used long before customers begin to become marginal. The best time to attack attrition is when someone first becomes a customer. Seek repeat purchases, increased average orders and purchases from additional product categories quickly while new customers' patronage is fresh. For example, an offer that accompanies a welcome to a new customer might easily perform 10 times better than a "we want you back" offer sent after the customer has become marginal. It would be a shame for customers to receive a win-back offer without ever having received a welcome offer!
All customers have some likelihood of responding to any offer. The challenge is to time the offers and to find the most productive and actionable segmentation. Waiting to create win-back offers for customer segments about to drop below breakeven profitability is action that's too little, too late.
STRATEGIES MUST BE TESTED AND TRACKED
An attrition model can be used to formulate an initial strategy, but once the marketing efforts start to change things, testing is crucial. It is unlikely that customers who spent a small amount two years ago will suddenly respond better than recent buyers of big-ticket items. However, they may behave differently if given targeted offers geared toward their segments.
Why? For some segments, increasing average order is important and offers should be created accordingly. For others, response might be low and offers should be geared towards getting another sale. Some other segments respond to offers that are non-seasonal while others may wait for a seasonal offer.
Testing and tracking new results will enable you to learn from the new, segmented offers. This knowledge will build upon the insights provided by the Attrition Analysis model and allow you to achieve even better results over time.
Questions? Just ask! [Back to How-to Resources] [Top] |