How matching works

The five questions

Everything starts from five questions about how your learner works and how much of you there is to go around. Every option is something you can observe at your own table — nothing needs a diagnosis, a report, or a vocabulary you have to learn first. These are the five, exactly as the app asks them:

• What does most of the learning?
• How long before they're done?
• Busy, colourful pages — help or hindrance?
• Is reading the hard part, or the point?
• How much can you be there?

What happens to your answers

Your answers are saved to your learner's record, so they are there when you come back — the privacy page lists them under "What we store". The comparison itself runs in your browser: the app downloads a copy of the catalogue and ranks it on your device. Matching sends nothing extra anywhere, and if you narrow the results by subject, that filter lives in your browser too and is never written down — which subjects a family is worried about is not a record we keep.

How the shortlist is made

Two stages, and they are different on purpose. First, hard rules remove what cannot work, and each removal keeps its reason: a resource offered only in some countries — removed with exactly that reason when we do not know where you are, never a guess — one that needs an adult working alongside when you said it has to run alone, one built around a timer. Second, everything left is scored against your answers using a small, fixed set of weights — the learning-style, session-length and reading questions count slightly more than the others. The checks are simple declared facts: does it have audio, is it hands-on, does it run in chunks your learner can finish, are its pages plain. No machine learning ranks anything here — the same answers give the same list every time, ties break alphabetically, and that is deliberate: a ranking we cannot explain to you is one we will not run.

You get 3 results at most, and sometimes fewer: a resource that matches nothing you told us is not a match, and an honest 2 beats a padded 3. Every result says why it is there — the reason is built from the same checks that scored it. And when you said it has to run alone but the seller has not said whether it can, the result carries that note — a fact about our information, never a mark against theirs.

What is never allowed to touch the ranking

No diagnosis — there is no field for one, and that is a matching decision as much as a privacy one: two learners with the same label routinely need opposite things, so we match on what works, never on what a report says. No age. No popularity — a resource is never ranked higher because other families picked it, since that loop feeds itself. No payment: the listing fee pays for a person to review a submission, it does not buy placement, and a paid submission can still be declined. And no ownership: One resource in this index is made by the same team that makes NeuroFlow. It is ranked by the same rules as everything else — nothing gives it a boost, a badge or a better position, and who owns a resource plays no part in where it appears. We tell you when you are about to open it, because you should know who you are dealing with, not because it changes where it appears.

What we check, and what we do not

A person reads every submission before it appears, and your link is checked before a listing goes live. What a seller declares about their resource stays attributed to that seller — we surface what they said and what you told us; we do not certify that a resource will suit your learner, and we never claim to have checked what is on the other end of an outbound link. The judgement that matters most stays where it belongs: with you.