I've been in this business a long time, and I still find many marketers focusing on the response rate as the primary driver for evaluating their campaigns.
“If we get a slam-bang response rate on this upcoming mailing,” one manager recently observed, “then we'll meet over half of our department's goals. Imagine, getting all those new customers with just one program.”
Don't misunderstand. There is nothing wrong with customer acquisition. However, as we approach the millennium, it's probably time that marketers begin to consider additional parameters as they make recommendations that directly affect the bottom line.
Take for example the banker who is offering a check cashing service. A $2 fee is collected every time a customer uses this feature, and a solicitation offering the service produces a 4.1 percent response rate. Make no mistake about it, this fee service — in addition to many others — is responsible for many banks reporting record earnings over the past years. However, this “cash-cow” profit producer is only worthwhile for those customers who will use it. With this in mind, does it really make sense to evaluate the program strictly on response? How about the individual who responds to anything that's free? Should he receive a solicitation at all? Responding here does not necessarily mean profit.
Credit card acquisition is the business of many banks and nonbanks. The current spread between what a credit grantor charges in the form of finance charge income and the interest the credit grantor himself pays to borrow money is huge. This is clearly a lucrative business if you manage your risk appropriately. For those firms that issue “fee-free” cards, the primary source of revenue is the interest charge assessed. Therefore, is it fair to evaluate the performance of a credit card solicitation with “response” only? How about spending? Even more important, how about balances that aren't paid off each month. These “revolving” balances are the key to a profitable program.
Attrition management is another area of concern. “Let's save those customers,” is the retailer's battle cry. After all, losing a customer means lost sales; so many retailers will develop loyalty programs designed to bond with the customer. Attrition models help identify potential defectors, and programs are designed to maintain the customer relationship.
This is all fine. After all, studies have demonstrated that a 5 percent reduction in attrition can significantly improve the bottom line. This often-quoted statistic is only true when those who are being saved are the ones that are spending. Sounds obvious, but you would be surprised at the number of marketers that gauge success by the number of customers saved and not by the amount of sales made.
Clearly, evaluating one piece of the puzzle is in many cases not satisfactory. In order to consider more than one part of the maze, multiple models are usually developed. Approaches to the above problems are typically handled by constructing two models:
* A probability-oriented model.
* A performance-oriented model.
In this example, a response model is developed. Added to this is a performance measurement model. So, at the end of the day, a manager might select those likely to respond and likely to spend. A strategy that may be pursued is to communicate to high responder and high spending cells.
Another way of viewing the picture is to mathematically “combine” response and spending. This is usually done via the expected value calculation. Let's demonstrate this.
The expected value is calculated by multiplying the probability of response by the spending prediction. A mailing program might consider one cell containing the highest expected values. Another cell might be designed to include response only. Some customers can be “coaxed” via good marketing, to spend more.
If we set our goals to identify the best spending segments, then why do we need two segmentation schemes? The answer is we don't. Targeting the high spending prospects is really all we need. And this was done with only one model. This is a complicated question that will be addressed in a future article. Suffice it to say for now, two schools of thought do exist.
One group maintains that only one model would be in order. The second group postulates that developing two models provides additional statistical rigor to the model's results.
The first model identifies responders. Clearly, spending can only emanate from responders. So, the second model operates only off the responder universe. Either way we look at it, response alone can't guarantee program success. It is only the fusion of good response coupled with good performance that may lead to a marketing victory.
Sam Koslowsky is vice president of strategic analytics of the marketing analytics unit at Harte-Hanks Direct Marketing, New York.