A finding nobody can check is an opinion.

Why the product is shaped this way

Most research ends up as a deck. A slide says seven of twelve participants didn't trust the saved card, and the room takes the researcher's word for it, because the clips are in one tool, the survey in a second and the tree test in a third. When the client's product lead asks “how do you know?”, the answer is an afternoon of digging.

So Stroop runs every method in one study, on the same participants, and keeps the evidence attached to the finding. A theme lists the clips behind it and the participant who disagreed. A sentence in a summary ends with its source, [P07 11:48], and clicking it plays that moment.

Numbers get the same treatment. The pietree, the similarity matrix, the SUS and the D-score are computed the way the specialist tools compute them, and the method is published, so an analyst on the client side can redo the sum.

Five commitments that cost us something

Every method published

How Stroop scores a D-score, a SUS, a tree test and a card sort, and how it times a response, is written up in the method notes with the papers it follows. Every stored score records the version of the code that computed it, and a note changes in the same release as its formula.

Every claim cited

The assistant may only say what a session, an answer or a diary entry shows, and each sentence it writes about participants cites the moment behind it. An answer with an uncited claim is rejected and written again before you see it. It never overrides a measured result, and when it can't find evidence it says so.

No training on your data

We never use your recordings, transcripts or your participants' answers to train AI models. Each workspace has one switch for AI, and studies marked sensitive start with it off.

Never recording without consent

Every recording needs the participant's consent for that study first, stored against the exact version of the form they read. A study with a recording task can't launch until it has a consent step.

Never running research for anyone's clients

We don't run studies, recruit or consult. So we never bid against the agencies that use Stroop, and a client's data is never ours to use for anything else.

Who runs it

Stroop is made and run by Never Before Seen Ventures Limited, registered with Companies House under number 16693370, at 41 Iffley Road, Hammersmith, London W6 0PB.

Email [email protected] and a person replies. There's no sales team: every price is published, and nothing needs a call.

What we won't build

Not a recruitment service

We don't find participants, and we never sell anyone's data. If you need people found for you, UserTesting, Maze and Lyssna run panels.

Not a lab

New paradigms and custom experiment code belong in Gorilla, Inquisit or PsychoPy.

Not native app testing

There's no SDK. On phones, participants use their own screen recorder and upload the file. And there's no SOC 2 report yet.

Not a research service

We build the platform. The research is yours.

Not AI-first

The assistant analyses and cites. It doesn't design your research or stand in for you.