Targeting using improved affluence data
Estimated income and other measures of affluence are often a driving force behind effective direct marketing campaigns. So what happens when your measures of affluence, namely income and home value, are simply inaccurate?
In order to produce comprehensive coverage of households at either a local or national campaign level, estimated incomes typically have to be derived from predictive models because there are only a few comprehensive sources for actual incomes - like the IRS or your bank. Validation results of current income estimates in compiled files are doing well if they are accurate 60 percent of the time within a band of ten thousand dollars.
For privacy reasons, these models are typically based on a limited number of self-reported incomes (in the thousands). Consequently, the results of these models can be, at best, rudimentary or directional. This can leave direct marketers with the unenviable challenge of producing the best results possible while having no choice but to base their list planning and targeting on fundamentally inaccurate information.
Virtually all estimated income models with national coverage are built from a limited sample of actual self-reported incomes and other key affluence measurements such as home values. The limitations of this self-reported data can easily lead to skews based on geography and cost of living. The limited size of these samples in combination with the application of a single national model means key factors such as higher or lower cost of living in certain locations or regions have to be ignored.
This result is a general degradation, or homogenization, of results across 120 million households. The “averaging of averages” effect within income or other affluence models causes households with low incomes to be overvalued and households with higher incomes to be undervalued. The bottom line: marketers who make assumptions about the demographics of which households to target based on affluence could easily be missing the actual target by 20-30% or more.
Because of technology advances and the introduction of new predictive modeling techniques, marketers today can have access to more accurate household income and discretionary income models based on the impact of local cost of living and demographic factors. What was once only available as a single national estimate can now be propagated as tens of thousands of individual models through automation. The models are derived separately and the results are validated and stitched together in a manner that addresses regional influences and cost of living weighting factors. The results of these models means a more accurate baseline around affluence and tool and a vastly improved targeting methodology and a more accurate result.
Marketers need to be aware of the flaws in traditional affluence models and know that with today's available resources they can obtain improved precision in their targeting of perspective customers.
Ray Kingman is CEO of Genalytics. Reach him at firstname.lastname@example.org.