Definition
In the product, the study you set up is called a Rehearsal, which contains one or more simulations. A simulation presents a group of replicas called a cohort with a scenario (a price change, an ad creative, a product choice) inside an environment (real-world context like macroeconomic conditions, local news, and product research for that replica’s area), and collects their responses.You create a Rehearsal in the intake chat and launch it from there. See
Starting a New Rehearsal for setup and
Running & Monitoring a Rehearsal for what happens after you
press go.
A simulation runs a moderated research interview against a cohort of replicas at once. Each
replica is interviewed independently, and their individual responses aggregate into a
distribution rather than a single averaged answer.
The Environment: A Moderated Interview
Every replica in a simulation is interviewed by an AI researcher in a natural, turn-by-turn conversation. The researcher asks your questions, listens, follows up to draw out genuine preference and reasoning, and when relevant shows the replica stimuli (images, video, documents) during the interview, the way a moderator shares a screen in a live study.The replica enters the scenario
The replica is given any environmental context for the study in its working memory, the
real-world backdrop the person would be reasoning against. For example, the current economic
climate, relevant market news, and the product research the person would do before deciding. The scenario itself is then put to them as a question by the researcher, so
nothing about the person’s own situation is assumed.
The researcher conducts the interview
An AI researcher asks your questions conversationally, adapting and probing for the real
reasoning behind an answer instead of accepting a one-word response.
Stimuli are shown when relevant
If your simulation includes creative or product artifacts, the researcher presents them during
the conversation and asks the replica to react.
Responses are captured
Each replica’s full answers and reasoning are recorded as a transcript.
Responses are open-ended and verbal. Replicas explain their thinking in their own words rather
than ticking a box. Structured outcomes, like which feature or ad won, are then derived by aggregating
across all the replicas in the cohort.
Anatomy of a Simulation
How Many Replicas?
By default a simulation interviews about 75 replicas per run. This number is adjustable, and may also fluctuate depending on how specific you make your cohort. Each simulation answers your full question set independently, so every question is answered by dozens of distinct people. Larger cohorts can be used to capture more of the variance in divisive or long-tail topics.What You Get Out
The immediate output is the set of interview transcripts, with every replica’s answers and reasoning. From there, findings are aggregated into a project report: an executive summary, recommendations, a participant table describing the cohort, behavioral segments, surprising findings, and the methodology (sample size, cohort description, number of simulations and questions, and date range). See Running & Monitoring a Rehearsal for how the run progresses and where the report shows up.Next: Simulation Types & Templates
Pricing, product preference, advertising, and the templates you can start from.