Population-Level Incidence Rates vs. Panel-Level Guesses

Every sample provider has their own incidence rate tool. They're all terrible for the same reason: they only work with that provider's panel. You get a number that tells you what percentage of their panel matches your target, not what percentage of the actual U.S. population does.

That distinction matters. If you're quoting a client on feasibility, you need population-level incidence. The panel-level number is useful for operations, but it's not what goes into your proposal.

So I built one that uses real Census data. The American Community Survey gives you actual population distributions across age, gender, ethnicity, income, education, employment status, marital status, household size, and about seven other dimensions. I pulled this for 200+ Metropolitan Statistical Areas because most research targets specific metros or regions, not the entire U.S.

The math is straightforward: multiply the proportions for each criterion together, assuming independence. That assumption breaks down when your criteria correlate (income and education, for example), but it gives you a defensible starting estimate. It's the same approach most sample companies use internally, they just don't show you the work.

How it actually works

Select your demographic criteria, pick your MSAs or go national, and the tool multiplies through the proportions. You get a topline incidence rate plus a breakdown showing which criterion is narrowing your audience the most. Nine times out of ten, it's a behavioral or attitudinal screener that kills feasibility, but at least with this tool you can see how much of your audience you've already lost before you even get to those questions.

I use this in proposals constantly. Client says "we want women 25-34 in the top 10 DMAs who own a home" and within 30 seconds I can tell them their population IR is roughly 3.2% before we add any category-specific screening. That number sets the tone for the entire cost conversation.

Tech stack

React 18.2 with Babel CDN for the frontend. No build step, no webpack config, no node_modules folder the size of a small galaxy. The Census ACS data is pre-processed and embedded. Runs entirely in the browser. I use the same stack across most of these tools because it keeps deployment dead simple: one HTML file, drop it on the server, done.