DM News Essential Guide to Lists and Databases: Modeling Uses Compiled Files to Their Fullest
Compiled files actually got their start in the 1840s. A business information company employed Abraham Lincoln as a correspondent, collecting information about business data for use in business report compilation. Consumer list compilation began in Nevada, IA, before 1920 with the collection of auto registration information on index cards.
Mail-order catalogs and magazine publishing proliferated after World War II. Mail-order buyers and magazine subscribers became a new revenue stream for these industries, and mailing lists became the backbone of the list industry economy. By the early 1960s, mail-order buyer and subscriber files had eclipsed compiled file mailing results. Consumers sourced from these files had a known penchant for ordering/responding by mail. Therefore, they were more apt to make purchases or respond to a specific mail offer.
The list industry was thriving on transactional information, highly correlated to various types of behavior required by mail-order businesses, magazine publishers, nonprofits, financial services companies and the like. Meanwhile, compiled files became hugely successful in the packaged goods space.
It was not until two large consumer list mailers experimented with regression modeling in the early ?60s that compiled files were ?retooled? for mail applications. Using the statistical technique of multivariate regression, these mailers were the new ?data miners? of their day, proving the value of a model?s priority household ranking based on predicted behavior, be it response, approval, number of shipments, donation amount, etc.
During the next 40 years, DMers were challenged to find the right balance within their mailing universe between two diverse list types: highly responsive, but more limited in quantity, or compiled ? ranked high to low based on a modeled behavior, offering the larger universes.
Fast forward to 2005. The cat is out of the bag: Regression-modeled compiled file names have become a standard component of most DMers? mailing list repertoire. For example, The Canine Fence Co., a Wilton, CT-based distributor of the Invisible Fence product, has increased its customer acquisition rate since it began testing a regression-scored compiled file.
Before list modeling, the firm relied on its customer profile, which naturally included a pet owner indicator to define its targeting strategy. But when this empirical approach began to show diminishing results, a new approach was required. Canine Fence tested a regression-scored compiled file. Test mailings proved that a statistical approach resulted in a higher acquisition rate over its empirical approach.
The company now has the best of both worlds: reliance on its original target market definition as well as a predefined, prequalified universe derived from statistical selection.