Looking at non-responders and eliminating them from a future targeted marketing campaign seems to be a “best practice.” It makes logical sense: If you get a 10% response rate, that means 90% of your effort was wasted. Predicting who will be in that 90% next time and eliminating them makes you more efficient. Right? And better analytics and more history means you will get better and better at dropping those target names. I’m sure most marketers will agree.
Before you begin cutting down your target list by 90%, let me play the devil’s advocate. Although the above statements are true, tthere are some important factors that work in direct opposition to them, and they need to be factored in as well.
1. As your knowledge grows and you become more proficient at identifying who is not responding, you could also correspondingly increase your knowledge of how to better define and allocate offers. This means that some of the people who didn’t respond in the past could respond to improved offer targeting in the future.
2. Just as our targeting will never be 100%, our identification/prediction of who will not respond is also never 100%. But the economics of these two types of error are not equal. Here’s an example. Say you have a postcard mailing to a retail customer list, with total production costs per piece averaging thirty cents. Communicating to someone who doesn’t respond costs you about thirty cents. That adds up to a lot when you are targeting millions of people, but how much incremental sales do you get from responders (this depends on purchase frequency and average order size for your products and services)? You can often get one incremental trip because of a response to a direct mail piece.
When you eliminate non-responders from your target list, you will invariably miss some of these folks, too. And in this example, for every 100 targets eliminated, if you miss even one of these responders, your production savings would be completely offset by a loss of incremental sales.
In order to be worthwhile, your cuts would need to be performed with better than 99% certainty that you are not eliminating any would-be responders. For media with lower costs per recipient -- e-mail, register receipt, apps -- the ratio could be even more dramatic.
3. When you reduce targets you often increase unit production costs. This is most significant in direct mail. Printing costs per unit go up. Postage can go up as well, because there is a discount for heavier saturation. Eliminating people in one postal carrier route can make that route reach a tipping point where all households on that route cost $0.01, $0.03, even $0.05 more each. These increases can partially offset the savings you seek from eliminating non-responders.
Don’t get me wrong. I love the idea of modeling behavior to see how people interact with offers and media. Response modeling is a good idea. But I include the cautionary advice above because I've seen that when marketers do not account for these things, then work very hard only to produce the opposite of their desired result.
A Samurai sword is superior because of its balance of strength and sharpness. These things are in opposition: A sword sharpened too much has a very weak edge. A strong edge can only be so sharp. When using CRM data for targeted marketing, our goal should not be to continually sharpen, but rather to continue to improve our ability to find the right degree of sharpness.