Concept test

Concept testing is market research that puts an idea still on paper in front of the people it is meant for, before anything is built. Most often it is product concept testing, but the idea can also be a service, a name or a promise. The test measures whether it is understood, believed, relevant, different and wanted, so that weak ideas are dropped or improved while changing course is still cheap.

A concept test happens early. That is its value and its limit: it steers a decision about what to develop, it does not predict what will sell.

The essentials

• A concept test compares ideas and explains the gaps between them. Its best output is a ranking with reasons, not a single score.
• Synthetic research changes the scale: testing thirty concepts instead of three maps the options before real measurement is spent.
• It is not a go or no-go. Saying and buying are different acts, for people and for models.
• The natural next step is product concept optimisation: rewriting the concept on the diagnosis, then retesting it.

What does a concept test measure?

The core battery is stable across the industry. Comprehension: is the idea understood as intended? Relevance: does it answer a need the audience recognises? Credibility: is the promise believed? Uniqueness: is it different from what already exists? Appeal and purchase intent. Often, value for money at a stated price. Open-ended answers add the first objection, which is usually the most useful line in the report.

What are the main concept testing methods?

• Monadic. Each respondent sees one concept. It is the cleanest comparison, and every concept needs its own sample.
• Sequential monadic. Each respondent sees several concepts in rotated order. It is cheaper, with some order effects.
• Concept screening. Many rough ideas are rated quickly to shortlist a few for a fuller test.
• Qualitative testing. Interviews or groups explain why a concept works or fails. They are good for improving a concept, and weak for choosing between close ones.

How is concept testing different from usability testing?

Concept testing asks whether an idea is worth building. Usability testing asks whether something already designed can be used. The first comes before development, the second during or after it.

Can concept testing be done with synthetic respondents?

Yes, and this is where synthetic research genuinely moves something. The bottleneck in conventional concept testing was never the method, it was how many ideas you could afford to test. Going from three to thirty does not make each test better; it lets you map the space before deciding where to spend real measurement.

What reads: the ranking of concepts, the gaps between them, where comprehension breaks, the objections. What does not read: a norm-based purchase-intent score, or a forecast of sales. A concept can land well with synthetic respondents and die on shelf, because the gap between saying and buying exists in humans too, and a model has never faced the arithmetic of the last week of the month.

Rankings depend on spread. If simulated respondents agree with each other far more than real people do, the gaps between concepts shrink or swing from one run to the next. We measured this under-dispersion in our own system against national surveys in eight markets: 0.20 to 0.76 of the human spread before correction, 0.96 to 1.07 on the attitudinal layer and 0.85 to 1.00 on the personality layer after it, where 1.00 means human-equivalent. The figures are in How reliable are synthetic respondents?

What comes after a concept test?

Concept optimisation. The test says which concept is ahead and what holds the others back. Optimisation rewrites the concept on that diagnosis, one change at a time, and retests the new versions against the original on the same audience. With synthetic respondents, a round costs hours rather than weeks, so the loop can run until the gain stops.

What can a buyer check?

• That concepts were rotated rather than always shown in the same order.
• That the audience matches the market being entered.
• That the report keeps what was ranked clearly apart from what was measured.
• That the ranking survives a second, independent run, and that the spread of answers is compared with a human reference.

When the idea to test is a name, see name testing.