Extrapolation & Census Data

Meta’s advertising ecosystem covers roughly 60–70% of the US population. When Consumr.AI uses Meta audience data as a behavioral signal source, the numbers returned represent that subset, not the broader population.

The denominator mismatch problem

Suppose you query a cohort on Meta and get back an audience size of 70 million people. But the actual US population matching that demographic description is 150 million. The denominators do not match.

If you use the Meta figure as your base and extrapolate from there, you will produce estimates that systematically misrepresent the full population, for example skewing toward the demographics that are overrepresented on Meta relative to the real-world population.

The same mismatch applies to demographic breakdowns within an audience. The age, gender, and income distribution of a Meta cohort may look quite different from what the US Census says that demographic actually looks like.

The fix: ACS/census distribution

The American Community Survey (ACS) is the US Census Bureau’s primary tool for collecting detailed demographic data between decennial censuses. It provides population estimates broken down by age, gender, income, geography, and other characteristics.

Consumr.AI uses ACS data to establish the target demographic distribution for a respondent pool. Instead of letting the Meta distribution dictate who shows up in a survey, the platform matches the respondent composition to what the census says the population actually looks like. This ensures:

• Age bands are represented proportionally (not skewed toward age groups that are over-represented on Meta)
• Gender splits reflect the actual population
• Income brackets match the real distribution

When this applies: respondents, not segment-building

Census-matched extrapolation applies specifically to respondent creation for surveys; it does not apply when building segments or audience segments for behavior reports.

When you are building an AI twin from a Meta-sourced audience, you use the Meta data as-is; the audience is a real behavioral signal and the Meta distribution is the correct base. When you are constructing respondents who will answer a survey and whose results need to represent the broader population, you apply the census correction.

The rule of thumb: if you are asking “who is in my Meta audience?”, use Meta data. If you are asking “how does my survey represent the full population?”, use census data.

How it works in practice

When you create respondents for a survey in Consumr.AI, the platform looks at the ACS demographic distribution for your target population (national or segment-defined) and constructs the respondent pool to mirror that distribution. The platform then runs a basic survey to establish demographic splits and segments based on that census-aligned foundation.

The result is a respondent pool whose demographics align with the real population, whose survey responses can be scaled up to population-level estimates without the denominator mismatch that would otherwise distort results.

Connection to other concepts

• Core Concepts → Respondents & Extrapolation covers how respondent pools are built and managed.
• Sampling & Sample Size explains the population-size benchmarks that determine how many respondents you need.
• Design Effect (DEFF) addresses how diverse respondent sourcing keeps the clustering penalty low.