FlashInsight

Glossary

Synthetic survey

Also called simulated survey · synthetic quantitative study

A conventional quantitative protocol — closed questions, segments, weighting — run against a population of synthetic respondents instead of a human panel.

A synthetic survey takes the standard quantitative apparatus — scales, closed questions, open-ends, segment crossings — and administers it to a population of synthetic respondents. This is the breadth route: several hundred profiles through the same protocol, in minutes.

Its value is not that it replaces a wave. It is that it changes the cost of trying. You move out of scarcity — three hypotheses tested because a wave is expensive — into something closer to experimental abundance: thirty explored before committing to one.

What it is not

Simulating is not measuring. A synthetic survey does not produce a market share estimate, nor a number you can put in front of a regulator. It produces differences between options, and differences are what you should read it for — never the absolute levels.

What a buyer can check

That the population is calibrated rather than improvised. That the spread of answers looks like humans and not like a model eager to agree. And that the deliverable says what it is: an instrument for steering, not a verdict.

Frequently asked questions

What is a synthetic survey?

A conventional quantitative protocol, with scales, closed questions, open-ends and segment crossings, administered to a population of synthetic respondents instead of a human panel. It is the breadth route: several hundred profiles through the same questionnaire.

What is a synthetic survey for?

Changing the cost of trying. Instead of testing three hypotheses because a wave is expensive, you can explore thirty before committing to one, then keep fieldwork for the most promising.

Can a synthetic survey replace a real survey?

No. Simulating is not measuring. A synthetic survey does not produce a market share estimate, nor a number you can put in front of a regulator. It produces differences between options, and that is what it should be read for, never the absolute levels.

See also

  • AI focus group — Several synthetic respondents brought together with a moderator, reacting to a stimulus and to each other. The depth route, on a small group.
  • Synthetic respondent — An instance of a model asked to answer questions as a human would, calibrated on data describing a real population. The output — the unit a sample is made of.
  • Under-dispersion — The commonest defect in simulated panels: answers cluster more tightly than real humans’. Disagreement disappears, and the signal goes with it.

Further reading

Back to the glossary