Pleasantville, NY: Pop. 30 Million

THE ABILITY to create prospect models from The Reader’s Digest Association Inc.’s file has opened the list up to a new set of direct marketers that now effectively has an additional 30 million names available.

The Reader’s Digest targeting services program uses about half the company’s customer and prospect file as the basis of its database analysis functions. While the Pleasantville, NY-based publisher has offered modeling services quietly since early 1999, it chose to keep the program under wraps until it could be validated by field tests.

The process starts with the 30 million individuals on the file who have seen and declined opt-out offers during the past year and are therefore available for targeting under Reader’s Digest’s privacy guidelines. Of those, 8 million have made a purchase from the company over the past 12 months. These people comprise the first level of prospects the Digest’s modeling department examines. The remaining 22 million older names are mined when the supply of appropriate prospects in the initial segment has been exhausted.

For modeling, the magazine’s file does not rely on overlay data. The models tend to stick to recency, frequency and monetary data rather than demographic and psychographic information.

The modeling process uses logistic regression, which, according to James Lynch, the Digest’s associate director for database marketing, can provide a numeric likelihood of matching a given desired behavior. Sometimes, when characteristics other than traditional RFM selects are paramount, the older names can yield more fruitful modeling information.

This was the case with apparel cataloger/retailer L.L. Bean of Freeport, ME, which had long resisted using the Reader’s Digest list. “[They said] `You are older, middle American, and we are younger. There’s no way we can slice and dice your names,'” says Diane Silverman, the Digest’s director of list sales.

But Silverman was convinced there were names for Bean in her company’s file, and made a point of urging the cataloger to test the modeling process.

In Bean’s case the model was based on a 200,000-name sample of its best customers. The best-customer list was run against Reader’s Digest’s 30-million-name universe. The programmers found a 40% match rate between Bean’s best customers and the Digest’s house file.

Bean’s model was built in April. The company will use the names of 300,000 recent purchasers, as well as 100,000 of the top scorers among the older 22-million-name file, and test them in this September’s Christmas catalog mailing.

One early adopter was Greenwich, CT-based business publisher Boardroom Inc. Boardroom’s initial tests of traditional selects from the Digest’s list were not encouraging: While order volume was satisfactory, pay-up and continuation rates didn’t justify further use.

If that wasn’t discouraging enough, Boardroom vice president of marketing Brian Kurtz believed that modeling will usually not make a loser list into a winner. Even with the best selects, he felt such processes will usually boost the top deciles only, and then just by around 40% to 50%.

But Kurtz was willing to gamble based on his knowledge of the file’s depth.

“My instinct said the selects they were offering [on list rentals] were so broad and non-segmentable, and that what was available was so much richer, that it was worth a test,” he says.

Boardroom’s 200,000-name sample was modeled against only 8 million of the Digest’s recent purchasers. At the time of the test, Reader’s Digest was still implementing the targeting services process, determining how to bring samples in and run them through the system, and was not running models against all available names.

Kurtz used targeting services’ profiling techniques to generate 13,000 mailing prospects for Bottom Line Personal, his company’s flagship publication. The promotion he used was a one-year soft “bill me” solicitation.

While the front-end results were not spectacular – Kurtz estimates that the response rate was around 20% higher than Boardroom’s previous uses of the Reader’s Digest file – the pay-up rate was “significantly better.”

He notes, however, that at the same time Boardroom tried the modeling technique, the marketing department also started using a new billing method.

The best deciles from the Reader’s Digest file have performed as well as good continuation lists in terms of profit and loss, says Kurtz, who acknowledges that the file is still outperformed by his blue-chip list rentals.

Even so, Kurtz took advantage of the file’s breadth of names to mail a much larger continuation – several hundred thousand – in February.

Currently, this type of modeling makes up around 10% of Reader’s Digest’s list rental revenue. However, Silverman expects that figure will rise to between 20% and 30% within three years.