Candidates Use Predictive Analytics to Seek Votes - How microtargeting works (
Page 2 of 3 )
Customer segmentation is, in a general, a way to group people or
organizations with similar demographic profiles, attitudes, purchasing
patterns, buying behaviors or other attributes to help organizations understand
customers more thoroughly and ultimately market to them better. Companies like
Strategic Telemetry and SPIN (Southern Political Information Network), a
Southern states political consultancy, use the same concepts in the political arena.
Microtargeting uses data from a number of sources—individual states' voter
files and census data—and combines it with available marketing data from
companies like InfoUSA and Experian to get a basic voter picture of a given
geographic area. Strategic Telemetry uses SPSS'
Clementine data mining and predictive analytics workbench, combined with its own
proprietary algorithms, to search for the best ways to reach the most likely
undecided voters.
Based on an individual's ZIP code or background information, campaigns can
get a better sense of who the voters are in a given area, and what the right
messages would be on key issues such as education, health care, immigration and
the economy.
Last week's primary election in Ohio
is a good example. Both the Obama and Clinton campaigns went in emphasizing
messages about the economy, as a large number of Ohioans have lost their jobs
in recent years.
"We geocode the files and apply census block data," Strasma said.
"That gives us thousands and thousands of data points—income, age,
industry that people work in—and from that we can determine how many blue
collar versus white collar employees, how long commute times are, what
percentage are agricultural versus suburban, and their country of origin. It
isn't about an individual—you don't look up if Joe Smith has a particular
income—you find out a particular [census] block income level."
Once data input is determined, companies like Strategic Telemetry utilize
algorithms to model various scenarios. Strasma said while the models can vary,
the three most common search targets are: undecided voters "because you
want your persuasion message going to the right people," the right
supporters to turn out for campaigns because "if you want to turn out a
voter you want to be reasonably sure they are going to vote for your candidate,"
and knowing which issues voters care about.