If you spend anytime in direct marketing, you’re going to run into a little problem. Let’s say you have a promotion that pulls in new customers at a profit — or so you think. Then fast forward six months or a year later and you find yourself with heaps of customers who responded once, but never bought again. What happened?
This problem is caused by something called the 'back—end performance' of a list.
Typically, most direct marketers focus on 'front—end' performance measures, namely on getting a good response rate. Folks are less likely to look very carefully at how those customers performed over time. And this puts you on a collision course with one of the most frustrating aspects of direct marketing:
Front end response rates are usually inversely proportional to back—end performance. The easier it is to get a customer, the less likely they are to buy again. And vice versa.
Take this example, from a book club mailer. In this example, the mailer sends out a promotion offering respondents the chance to join a club allowing them buy one book a month for the next year, at a substantial discount on the retail price. On average, folks who join this book club typically buy 10 books over the following 12 months:
| List A | List B | List C |
|---|
| Response Rate | 1.00% | 1.50% | 2.00% |
| Responses (per 1,000 mailed) | 10 | 15 | 20 |
| Cost (per 1,000 mailed) | $800 | $800 | $800 |
| Cost/Sale | $80 | $53 | $40 |
| Avg # of books sold (next 12 mo.) | 10 | 10 | 10 |
| Profit per book sold | $5 | $5 | $5 |
| Profit per response (next 12 mo.) | $50 | $50 | $50 |
| Net Income / response | ($30) | ($3) | $10 |
Looking at these results, we would correctly conclude that only List C is profitable to mail. Then when we 'rollout' or re—mail List C, we might get this result:
| List C |
|---|
| Response Rate | 2.00% |
| Responses (per 1,000 mailed) | 20 |
| Cost (per 1,000 mailed) | $800 |
| Cost/Sale | $40 |
| Avg # of books sold (next 12 mo.) | 6 |
| Profit per book sold | $5 |
| Profit per response (next 12 mo.) | $30 |
| Net Income / response | ($10) |
Most customers buy 10 books, but for whatever reason, folks who join from List C tend to buy less. Only 6 books a year.
This is because not all lists are created equal. It can therefore be misleading to use the overall average number of books sold. The actual number of books sold, by list, can vary dramatically.
Going back to the original test of three lists, let's look a little more carefully at the back—end sales figures — by list:
| List A | List B | List C |
|---|
| Response Rate | 1.00% | 1.50% | 2.00% |
| Responses (per 1,000 mailed) | 10 | 15 | 20 |
| Cost (per 1,000 mailed) | $800 | $800 | $800 |
| Cost/Sale | $80 | $53 | $40 |
| Avg # of books sold (next 12 mo.) | 18 | 12 | 6 |
| Profit per book sold | $5 | $5 | $5 |
| Profit per response (next 12 mo.) | $90 | $60 | $30 |
| Net Income / response | $10 | $7 | ($10) |
As you can see, customers you acquire from lists A and B come in at a lower response rate, but their back—end performance is significantly better than list C. And both lists A & B are actually profitable to mail. List C is not.
Still, being that we'd like to mail as many lists as we can, we shouldn't just discard List C as an option. Because, embedded within most lists are profitable segments. We just have to find them.
Everyday, direct marketers struggle with lists that don't work well, or don’t work as well as we would like. The best way to work—around that problem is through segmentation, the process of slicing up a list by a predictive attribute that will boost response rate enough to turn a profit.
There are three common reasons why a list won’t work, and three common segmentation solutions for each problem.
Problem #1 — Response rate is too low. The best way to segment your way out of this problem is to focus on recency of last purchase. Folks who bought something from you less than 3 months ago will respond at a high rate than folks who haven’t bought anything in 4 months or more.
Problem #2 — Average purchase is too low. Your best bet is to segment the file by prior purchase amount. Someone who spent $100 in one go is far more likely to make another big purchase. Folks spending less than $10 in the past will most likely continue to spend frugally.
Problem #3 — Customer retention is too low. This is the problem we face with the book club above. Our customers are not staying with us over the long—run. The best way to solve this headache is to focus on frequency of past purchases. Someone who has already bought from you twice is far more likely to become a regular customer, vs. someone who has only bought once from you.
So let’s now look at List C once again. Let’s say the books you are selling all pertain to DIY home renovation and decorating. And let’s say that List C is a list of subscribers to a gardening magazine that your company also publishes.
Segment List C by tenure. Meaning, sort out the folks who bought their subscriptions this year from the folks who have subscribed for 2 years or more:
| List C |
|---|
| All subs | 1 year (or less) | 2+ years |
| Response Rate | 2.00% | 1.00% | 3.00% |
| Response/M | 20 | 10 | 30 |
| Cost/M | $800 | $800 | $800 |
| Cost/Sale | $40 | $80 | $27 |
| Avg # of books bought | 6 | 3 | 9 |
| Value per book sold | $5 | $5 | $5 |
| Total Value | $30 | $15 | $45 |
| Net Income/Response | ($10) | ($65) | $18 |
In this example, not only do the 2+ year subscribers buy more books, their response rate is also higher, which takes you from losing $10 per sale to making $18 per sale.
You’ve just stumbled upon one of the great little 'secrets' of direct marketing:
When segmenting your file by frequency of prior sales, you get to have your cake and eat it too: Response rates and retention both rise.
I’ve worked on a wide range of direct marketing promotions over the years, in the US, Europe and here in Australia. And everywhere, this problem — and its solution — is the same.
Company size and industry matter little. Giant telecoms often acquire massive numbers of new customers, only to find them “churning” out 6 months or a year later. While across town, a small charity will struggle with a list of “tips”, previous donors who gave $10 once and never give again. It’s the same basic problem for both.
Back—end analysis can solve this problem for you, and it takes far less time than one might imagine. You rarely have to wait a full year to complete these analyses. Most back—end measures follow a “pattern”, which can usually be measured and projected with just a few weeks or months worth of data.
12 years ago, I made the mistake above — and then learned how to solve it in order to save my job. I succeeded, but it would be great if you didn’t have to learn this lesson the same way. So beware, and best of luck!