Writing Eligibility Criteria That Don't Kill Your Response Rate
Every extra qualifying condition shrinks your reachable pool. Here's how to tell which criteria are worth that cost.
Every eligibility condition you add — age range, income band, a specific behavior, a category interest — multiplies against every other condition, not adds to it. Two criteria that each match 50% of your pool don't leave you 50%; combined, they leave you closer to 25%. This compounding effect is the single most common reason a study stalls, and it's invisible until you actually look at the math.
Start from the question you need answered, not the description of your ideal customer
It's tempting to write eligibility criteria as a full customer profile — age, income, location, brand affinity, purchase frequency, all at once. But most research questions only actually depend on one or two of those conditions being true. Ask which conditions the analysis genuinely breaks without, and treat the rest as optional segmentation you can capture as a *question inside the study* instead of a *gate before it*.
Order your criteria from broadest to narrowest
If a screener has to check five conditions, checking the narrowest one first means you pay the cost of showing the study to everyone before rejecting most of them late. Ordering broad-to-narrow doesn't change who ultimately qualifies, but it does mean the criteria most likely to disqualify someone are evaluated last, after the ones that are cheap to satisfy — this keeps your visible acceptance rate readable instead of front-loaded with rejections.
Distinguish "must be true" from "nice to know"
A criterion belongs in the eligibility screener only if a response from someone who doesn't meet it would be *unusable* to you, not just less interesting. Everything else — income bracket for a study that isn't income-sensitive, a secondary product preference, general demographic color — belongs in the study itself as an early question, where it costs you nothing in reach and still gives you the segmentation data afterward.
Re-check criteria against actual completion data, not assumptions
A criterion that seemed necessary at study-design time sometimes turns out not to correlate with anything in the resulting data. If a criterion isn't producing a measurable difference in your responses after the first batch of completions, it was gatekeeping reach for no analytical benefit. The [Study Detail page](/sphere/studies) shows screen-out counts per study; a criterion responsible for an outsized share of screen-outs relative to what it buys you analytically is a strong candidate to relax.