This is the most expensive misunderstanding in the category. A brand runs a fast, inexpensive in-home test, gets an encouraging percentage back, and puts it on the pack. The number is real. The study was real. The claim is still not supportable — and nobody finds out until legal review, or a retailer, or a competitor asks how it was produced.
Two different jobs
Consumer research does two quite different jobs, and they are easy to confuse because both produce percentages.
The first job is deciding what to fix. A small, fast study tells a formulation team whether the sweetness is wrong, whether the texture is the problem, whether people would use it again. It is an internal instrument. It is meant to change the product, not the packaging.
The second job is substantiating a sentence. That requires the study to have been designed in advance to answer one specific question, with the analysis fixed before the data arrived. It is an external instrument, and it is a materially different piece of work.
At curí the first sits at DISCOVER and the second at VALIDATE. The distinction is not a pricing tier. It is what the output is allowed to be used for.
Why a small fast study cannot carry a claim
Three reasons, none of which are about effort.
No prespecified endpoint. If the primary question is chosen after seeing the results, then across enough measured attributes something will look good by chance. Choosing afterwards is what makes the number unreliable, regardless of how large it is.
No analysis plan. Without a statistical analysis plan written before unblinding, decisions about who to include, how to handle dropouts and which comparison to report are all made while the answer is visible.
Not enough people, for long enough. A short test with a small cohort can be perfectly adequate for direction and quite inadequate for a confidence interval anyone should rely on.
None of this makes a discovery study bad. It makes it the wrong instrument for the job of substantiation — the way a thermometer is not a bad ruler.
What VALIDATE adds
A prospective protocol, written and fixed before recruitment. A prespecified primary endpoint — the single question the study exists to answer. Randomization, and blinding wherever the format allows it. A statistical analysis plan agreed in advance. Daily diaries where the endpoint is experiential. A formal report with complete statistical tables, and claim-support recommendations that state plainly which sentences the data will and will not carry.
That last item is the one brands underuse. The most valuable output of a well-run consumer study is often the list of claims it does not support, because that is the list that would otherwise have reached a lawyer.
The practical sequence
Run the cheap study first, but run it for the right reason. Use DISCOVER to make the product better and to find out whether the effect you hope to claim is plausibly there at all. If it is, design a VALIDATE study around the specific sentence you want — starting from the wording, working backwards to the endpoint.
Brands that do it in this order spend less overall, because the expensive study is only ever run on a product and a claim that have already shown they are worth it.
One boundary that does not move
All of this sits inside structure/function territory. Consumer evidence can support statements about how people experienced and accepted a product. It cannot be used to say a product treats, prevents or cures anything — no sample size makes a disease claim permissible for a dietary supplement.
This note describes curí’s own laboratory practice and is written for brand and regulatory teams. It is not regulatory advice on a specific product. All programs are designed to support structure/function claims permissible under DSHEA; curí does not design studies to support disease claims.