Transaction data drives direct mail industry.

AuthorLUSKY, MARK

Santa Claus could take a tip from Abacus Direct. The Broomfield-based statistical modeling company has a list, checked and updated on a regular basis, that tells clients much more than simply who's naughty and who's nice.

Abacus Direct has made information brokerage (e.g., list rental) efficient and profitable for those who use direct mail. Its key: using actual transaction data as the beginning point.

Abacus Alliance is Abacus Direct's proprietary database of consumer and business-to-business catalog buying behavior. Contributed to and used by more than 1600 client companies in the United States, Abacus Alliance provides those clients much more than names or even demographic information.

"The vast majority of the (data) market is list only. Abacus was the first to take a total view of the consumer and ... actually aggregate and model transaction histories across multiple companies," said Christopher Dice, president and chief operating officer of Abacus Direct since 1998.

It sounds perfectly logical and deceptively easy. But, according to Dice, when company founder Tony White first presented his idea to the catalog industry in 1990, the response was anything but enthusiastic. Traditionally, companies rent their proprietary lists one at a time to other companies. White wanted them to give him their lists to form one huge database, thus providing a much broader view of the consumer. Gradually, White convinced catalog merchandisers to cooperate.

To get information from the system, Abacus Alliance members first must contribute their own customer transaction information to the database. "To be a member, a company has to have customers first. We don't deal with start-ups," Dice said.

The Abacus Alliance database now includes 3.5 billion transactions representing 90 million households -- or virtually every household in the United States with discretionary income.

In addition to its database, Abacus early on developed a proprietary automated modeling process that it applies to the data. "We can create custom models," said Dice, "but we average 10,000 linear regression models monthly." Linear regression is a straight-line statistical model that forecasts the future based on the past. It minimizes errors between observed and predicted data. If, for example, the pattern is 2, 4, 6, and 8,10 is the...

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT