What is concept testing?

What it is

Show people a concept — a description, a storyboard, a landing page, a clickable prototype — and collect their reactions before committing engineering to it. The cost is days rather than quarters, which is the appeal.

The method is honest about being a proxy. The failure is treating the proxy as the thing.

What it reliably catches

Comprehension failures. Ask someone to explain the product back to you in their own words. If they cannot, no amount of demand would have saved the launch, because they cannot buy something they cannot describe.

Wrong mental model. People will tell you what they expect the product to do, and it is frequently not what you built it to do. This is cheap to learn now and brutal to learn after launch.

Category confusion. If half the room thinks it competes with a product you consider unrelated, your positioning has a problem that is much easier to fix on a landing page than in the market.

Obvious dealbreakers. Requires a bank login, needs my team to switch tools, only works on desktop — objections that arrive in the first five minutes and would have arrived in month six anyway.

What it cannot tell you

Whether anyone will actually use it.

Enthusiasm in a concept test is close to free. The respondent is not choosing between your idea and their current workflow, not paying, not migrating data, not persuading a colleague. Every one of those frictions is absent from the test and present in reality.

Quibi is the standard illustration: the concept — premium short-form video for phones, with real stars — tested and pitched extremely well. What testing could not surface was whether anyone would add another paid subscription for a format they were already getting free. Google Glass has a similar shape, where the concept was thrilling in a demo and socially impossible in a café.

Better signals, roughly in order of strength

1. Money. A pre-order, a deposit, a signed letter of intent. Costly to give, therefore meaningful.
2. A fake door. A real entry point in your live product leading to a not-yet-available page. Measures behaviour among people in context. Show the people who clicked an honest message and let them opt into being told at launch — anything else is a trick your users will remember.
3. Concierge delivery. Deliver the outcome manually for ten customers. Slow, unscalable, and the most reliable demand signal short of revenue.
4. Concept test. Useful, cheap, and the weakest of the four.

How to run one properly

Show the concept, then be quiet. Ask them to explain it back. Ask what it would replace and what they do today. Ask what would stop them.

Then ask for a small commitment — an email for early access, a calendar slot, a deposit — and record who actually gives it. That single behavioural question is worth more than every rating scale on the page, because it is the only moment in the test where the answer costs the respondent anything.