Compiled Lists - Take a Second Look (968 words)
Make It Work
Before you can sift through vast amounts of data, you must understand who your customers are. This requires a complete analysis of your housefile.
The degree to which you need to analyze your housefile, says Schneider, depends on the product or service sold.
For instance, the analytics involved in cloning newspaper subscribers is not as complex as those required of a marketer selling long-term care to senior citizens at a monthly premium of $700. In this example, says Schneider, "You're selling a niche product, and then analytics are key."
If this is the case, Schneider advises marketers to start by creating a customer clone model used to select records on a compiled file that shares the same characteristics as your customers.
Next, conduct a test mailing utilizing names from the compiled file. The response from this test mailing then can be used to build a logistic-regression model that looks at the percentage difference between the same variable on one file against another. The percentage of difference is used to detect and establish a segment that's likely to respond.
Schneider gives an example: If you mail 100,000 names from your housefile, and analysis of the responses finds that 90 percent of responders are homeowners, you'd use this model to select and mail an additional.100,000 names from a compiled file. Response analysis reveals that 99 percent of these responders also are homeowners.
As the percentage of responders to this mailing is equal or greater than that of the housefile mailing, the marketer then knows this is a segment likely to respond. Had the percentage of homeowners who responded to the mailing selected from the compiled data been, say, 50 percent, we'd have known that homeownership is not a response indicator.
But, as Dunhill is quick to point out, modeling isn't always practical for all mailers, particularly smaller volume mailers, which account for 90 percent of all businesses. For them, an alternative is to analyze their circulation lists and select the names on a compiled file that match those demographics and psychographics, suggests Dunhill.