One (persona, user_index) participant. 12 personas × 857 users = 10,284 independent samples, each judged individually.
A simulated respondent emits only what a person could: a choice, a stop step, free text, an ordinal certainty. Never a probability.
Rollups store counts frozen at finalize. Every rate, bound and gate is recomputed on read by one estimator, so what-ifs can't drift.
persona-summary → study-summary. A finding's strength is the number of distinct personas quoted, never a self-graded confidence.
Ask persona-based questions, scoring emotional and UX-dimension responses
Send personas through a real survey form and synthesize how they respond.
Compare two variants and see which one personas prefer — and why.
7 studies · last ran today · 12 personas