SuppleGut

How we check our results

The short version. The quiz is a set of written rules, not an AI. We ask the people who want to be asked whether the product they were matched to helped, and we publish what they tell us here once there is enough of it to mean something.

How the matching works

Every product in the catalog carries the same set of facts: which symptoms it is used for, what the human research shows for that exact formulation, what customers consistently report, how fully the label is disclosed, and any cautions for pregnancy, medication or particular conditions. Your answers first remove anything that conflicts with what you told us about your health, then rank what is left on those facts. The rules are the same for every product and every person, they are written down, and nothing in them changes for a brand. How we match walks through them in more detail.

How we find out whether it helped

Near the start of the quiz we ask how much your gut symptoms affect your day-to-day life, on a scale of 0 to 10. That number plays no part in the match; it is the starting point we compare against later. After your results you can ask for a check-in. If you do, we email you twice, two weeks and six weeks after the quiz, with one question: did it help? Each email has one-tap answers, and a short optional form behind them asks the same 0 to 10 question again, which product you actually tried, how consistently you took it, and whether anything else changed at the same time.

Every completed quiz is recorded under a random session id, with the answers tapped and the products shown. Check-in answers attach to that id, not to a name. Email addresses are kept in a separate table and used only to send the two check-ins. Our privacy policy sets this out in full.

What the numbers are, and are not

These are self-reported outcomes from people who chose to answer. They are not a clinical trial. There is no placebo group, nobody is blinded, and the people who reply are not a random sample of the people who took the quiz. What they can show is whether, for the people who tried a product the quiz recommended, symptoms moved in the right direction more often than not, and which products or categories that holds for.

How we filter

Bots and skims. A quiz finished in under twenty seconds is not a person reading the questions. Those sessions are kept but excluded from every statistic.
Low adherence. A product taken now and then, or stopped, cannot be judged on the result. Published figures use replies where the product was taken daily or most days.
Confounders. If someone also changed their diet, medication or other supplements, the change cannot be credited to the product. Those replies are reported separately.
Small numbers. A product's figure is published only once enough people have replied for it to mean anything, and the number of replies is always shown beside it.

Published results

Results will be published here once enough check-ins have come in: the share of people reporting improvement, by product and by category, with the number of replies behind each figure.

Average change in the 0 to 10 day-to-day score, from the quiz to the six-week check-in, for people who took the product daily or most days and changed nothing else.

Why people did not try the product they were matched to: price, availability, trust, or something else.

Aggregate figures like these may be published here and shared with others, including supplement brands and researchers. They are counts and averages; they never identify a person, and they never influence which product the quiz recommends.