This guide covers market surveys — not in-app user surveys, customer feedback tools, or post-purchase forms. A market survey measures populations using representative samples. It involves paying people to take your survey.

Part I covers how to define and size the population you want to sample. Part II covers the vendors you can use to recruit from that population.

Use verifiable metrics — age, gender, geography — that you can cross-reference with Census.gov, or any trusted data source. Generational labels (e.g., Gen Z, Millennials) don't count.

Once you've defined the demographics, look up the actual population count. For 18–34 year olds in the U.S., that's approximately 75 million people — derived from the Census by combining age brackets. Knowing the base population size anchors everything downstream: your sample size, margin of error, and quota structure.

Gender splits roughly 50/50, but the age distribution within 18–34 is uneven — about 40% fall in the 18–24 bracket and 60% in the 25–34 bracket. Set quotas based on actual composition, not intuition.

For strategic work — defining markets, sizing segments — I typically use n=1,100 to achieve a ±3% margin of error. Each dot below represents one respondent, split proportionally across the same four groups as the population in Step 3.

These are the primary sources I use. If you're based outside the U.S., reference your own national census equivalent.

One of the most common mistakes I see: using age brackets that don't align with U.S. Census brackets. If your brackets don't match, you can't compare your survey data to the broader population. Use the ones on the right.

Are you targeting a household decision or an individual one? For products used across a household — air purifiers, streaming services, cleaning supplies — think in households. For personal products — toothbrushes, skincare, supplements — think in individuals.

Scale reference: there are approximately 130 million households and 333 million individuals in the United States. That distinction will affect your incidence rate and your respondent fees.

If you need to target a niche segment — daily Vitamin C users, Tesla owners, fintech app adopters — Census data won't get you there directly. Search for existing industry reports and use them to estimate prevalence rates within the broader population. A few reliable sources:

Two variables drive most of the cost: Incidence Rate (IR) — how common your target respondent is relative to the general panel — and Length of Interview (LOI) — how long the survey takes to complete.

A high IR and short LOI means cheap, fast data. A niche segment with a 30-minute survey will cost significantly more per complete. Know your IR before you request a quote — vendors will ask for it, and a ballpark estimate helps you avoid sticker shock.

No matter which vendor you use, include at least one open-text question. Bots can't write coherent sentences. Respondents rushing through for the payment reward will write gibberish or copy-paste from nearby fields. A single open-text question lets you flag and remove low-quality responses before they contaminate your analysis.

Look for: duplicate responses, single characters, irrelevant content, or responses that have nothing to do with the question asked.

Some platforms charge per response rather than per respondent — which sounds reasonable until you do the math. Here's a real example from a client decision:

Vendors love to advertise exclusive, vetted respondent databases. Sometimes those panels are genuinely high-quality. More often, "proprietary panel" means a panel they own but supplement heavily with external sample partners — which is the same aggregated pool everyone else is drawing from.

Ask vendors directly: what percentage of a typical sample comes from your proprietary panel vs. external partners? The answer tells you a lot.