Why we said no to fake respondent counts
It would be easy to show '10,000 simulated respondents' on a panel result. We don't, because a panel of five distinct personas and a panel of ten thousand near-identical copies would look the same on that number.
A big number is the easiest thing in this category to fake, and the least useful thing to show.
It would be trivial to generate ten thousand synthetic responses instead of five and display a headline number that looks like statistical weight. It would also be actively misleading, because generating more copies of the same underlying model doesn't add information the way more real respondents would.
We made a deliberate choice not to lead with a respondent count, and this page is about why that specific number is the wrong thing to optimize for in synthetic research.
Why a big count feels credible
In real survey research, sample size is a genuine, meaningful lever — more real respondents narrows a confidence interval and reduces the chance a result is a fluke of who happened to answer.
That association runs deep, which is exactly why a large synthetic respondent count borrows credibility it hasn't earned: the underlying mechanism generating those responses is nothing like recruiting more real people.
Why it's misleading here
A synthetic panel's real limiting factor isn't count, it's diversity — five personas that genuinely disagree with each other the way a real population would tell you more than five thousand generated from prompts that collapse into near-identical answers.
Inflating the count without checking for that collapse produces a number that looks statistically serious while hiding the exact problem — mode collapse — that would make the underlying result unreliable.
What we show instead
A diversity score, so you know whether the panel you generated actually varies in its answers rather than repeating one opinion under different names.
An agreement score with an honest confidence interval, computed with a method (the Wilson score interval) that doesn't collapse to false certainty at small sample sizes the way the naive version would.
A sample-size calculator, so if you do want to know how large a real panel would need to be for a given precision, you get an honest answer grounded in real statistics — Cochran's formula — rather than a number we made up to sound rigorous.
The absence of that number is itself a claim: we'd rather show you a smaller, honest, checkable set of numbers than one large, impressive-looking one that doesn't mean what it implies.
The trade-off we accepted
Not showing a big count means thepanelist's results look less immediately impressive than some competitors' marketing pages. We accepted that trade explicitly, because the alternative — a number chosen for how it looks rather than what it means — is the exact kind of overstatement this whole site is trying not to make.
Common questions about respondent counts
Why doesn't thepanelist show a large simulated respondent number?
Because it would be easy to inflate and wouldn't reflect the thing that actually matters for a synthetic panel's usefulness: whether its personas genuinely disagree with each other, not how many copies were generated.
What should I look at instead of a respondent count?
The diversity check and agreement score for the specific panel you generated — both measure something real about that panel's internal consistency, unlike a headline count.
Does a small panel size mean the result is weak?
Not necessarily. A small, genuinely diverse panel is more informative than a large, collapsed one. Size alone was never the right thing to optimize for here.
See the real numbers behind a panel result
Diversity checks and confidence intervals, not an inflated headline count.
Read the methodologyMore Critical Perspectives
Synthetic panels are a gut-check, not a replacement
It would be easy to let the marketing copy drift toward 'skip user research entirely.' Here's exactly what a synthetic panel can and can't tell you, and why we keep saying so even when it costs us a cleaner pitch.
AI market research: how accurate is it really?
Accuracy isn't one number. It's a set of separate, checkable claims — and most of what gets called 'accuracy' in AI research marketing is actually a claim about something else entirely.
The honest limits of synthetic customer research
A running list of what synthetic research structurally can't do — not caveats buried in fine print, but the actual boundary of the method, stated plainly.