Beyond National Averages

Every audience tool on the market will give you age brackets and gender splits. Most will add income ranges and education levels. What almost none of them do is model how these traits co-occur.

That matters because demographics are deeply interdependent. Income depends on education, which depends on age, which depends on geography. A 22-year-old in Manhattan with a bachelor's degree has a completely different income distribution than a 22-year-old with a bachelor's degree in Topeka. Treating these as independent variables (pick an age, pick an income, pick an education level) produces profiles that look plausible on the surface but fail to correspond to anyone who actually exists.

PreFlight models joint demographic distributions at the city and county level. Instead of drawing each trait from its own national distribution, we sample from conditional distributions that preserve the real relationships between age, sex, education, income, employment, and occupation.¹ The result is a profile where every demographic trait is consistent with every other one, grounded in the actual population of the place you're targeting.

National averages hide enormous variation. The median household income in the United States is roughly $80,000. In San Jose, it's $140,000. In Brownsville, Texas, it's $40,000. That kind of variation is the norm in American economic geography.

PreFlight integrates demographic data at the city, county, and metro level across every state and territory.¹ Occupational distributions come from industry-level employment data that captures the actual economic structure of each region.² Religious affiliation is modeled using congregational membership data at the county level,³ calibrated against large-scale national surveys.⁴

The result is a demographic model where a profile generated for Portland, Oregon looks meaningfully different from one generated for Provo, Utah, because the underlying populations are meaningfully different.

Building realistic demographic profiles is fundamentally a problem of conditional probability. Rolling dice on each trait independently will produce nonsensical profiles. The real question is: given that this person is female, 35, and lives in this county, what is the realistic distribution of education levels? And given that education level, what is the realistic distribution of incomes? And given all of that, what industries are they likely to work in?

Each of these conditional relationships requires its own data pipeline, its own set of cross-tabulations, and its own validation against known population statistics. PreFlight chains these conditional distributions together so that the final profile is internally consistent from top to bottom. Political orientation is modeled using validated typology frameworks⁵ that go beyond simple party affiliation to capture the actual structure of American political identity.