Marketing modelers who use U.S. Census data better make sure their slide rules are properly oiled: This year marks the first in 70 that the Census Bureau won't mail a long-form questionnaire to one-sixth of the entire U.S. population. In its place, the Bureau is using the American Community Survey (ACS), an annual questionnaire.
There's a catch, of course: Unlike the long form, the ACS will be mailed to a comparatively small number of citizens, meaning the Census Bureau will rely on sampling, projections and weighting to a greater extent than ever before.
The good news is that the ACS will be sent out every year. “This may be an improvement, but it is a difference,” Don Hinman, Epsilon Targeting's senior vice president for data strategy observes, when thinking about how analysts will have to adjust their models. “The potential benefit of ongoing data is probably more useful than point-in-time data once every 10 years.”
“We're very excited about the ACS,” says Joan Naymark, director of market analytics and planning in Target Corp.'s capital finance department. “We're going to be able to watch changes over each decade in the characteristics of the communities in which we do business.”
Tracking the Boom
Naymark is especially interested in demographic shifts as generations mature. Take the baby boomers, who are moving into retirement age. Target will track where they will live, whether they continue to work, and if so, in what sorts of jobs.
Baby boomers aren't the only cohort Naymark is tracking. “How will the millennial generation organize their household formation years?” she asks. “Where will they live? The ‘where’ question is very important to targeting — it allows us to supply appropriate merchandise.”
There's a lot of money riding on the answers to Naymark's questions. Each new Target store represents a $20 million investment. Survey data also helps the company choose new outlet locations, aids decisions about remodeling and infrastructure, and influences merchandise mixes and marketing activities.
ACS data will be released in one-, three- and five-year aggregations, meaning that in 2016, marketers will be able to choose from 2015 ACS data alone, or combined results from either the 2011-2015 or 2013-2015 surveys.
The ACS is aggregating its data because its sample size for any given year is too small to allow granular analysis. As Ken Hodges, chief demographer at Nielsen Claritas notes, only 800 of the 3,141 counties in the U.S. are large enough that the ACS will provide a stable sample size for one-year data.
“There is a lot of concern about how much sampling error there will be for block groups, which are the smallest geographic units for which ACS information will be provided,” Hodges says.
That said, Hodges is very enthusiastic about the new system. “The real payoff is not that marketers can make a multi-million-dollar decision for a single block group,” he says. “The real payoff is that the data can be used in areas related to business applications. Businesses will define a 20-minute drive time around a given location. We can aggregate small-area data to the user-defined area. When you aggregate those data, the error rates come down rapidly.”
Not So Sure
While retailers are sanguine about the changes to Census data, at least one direct marketer has his doubts.
“In general, the more data we get from the Census, the safer we are,” says Dave Chambers of DaveTheWineMerchant.com.
For Chambers, the use of sampling and weighting to bridge holes in coverage is problematic.
“You are one-step away from what is actually found on the ground,” Chambers says. “And if you use those projections [in a model] and make further projections, then you are two-steps away.”
Chambers' own use of Census data focuses on comparing the demographics of his customers with those of the population in a specific area. “I did an analysis of first names to determine the gender breakdown of my customers. Most of my buyers were men, but most of my loyal buyers were women. By relating that to the marketplace, I could see how to better target my messaging, and where to go to find more.”
Network
