EvidenceField
Synthetic where data is rich. Human where it isn't.
AI personas promised to replace user research
The science says: only where research was already easy
The Evidence
Synthetic respondents flatten real people
Simulate 1,000 of our users.
Returned half the variance of real humans...
The averages look right — the relationships are wrong
What drives churn for segment B?
48% of synthetic coefficients diverge. Signs flip 1 in 3 times...
And accuracy falls with distance from the US
Now simulate drivers in Jakarta.
Similarity to real humans drops with cultural distance: r = −.70
Pure-synthetic research fails exactly where research is hardest
“
WEIRD in, WEIRD out.
So we built the model the science actually supports
Market-Adaptive Research
Synthetic panels where data is abundant — fast, cheap, exploratory
Human fieldwork where it isn't — real people, met where they are
AI does what it's actually great at
Synthesize 60 interviews across 6 languages.
Transcribing, translating, cross-referencing themes...
Experts route every study
New market entry: Vietnam. Design the study.
Routed: in-person fieldwork + synthetic pretest...
Every insight traceable to real humans
Show the evidence behind this recommendation.
14 field interviews, 2 synthetic pilots, 3 peer-reviewed sources
Built by an operator who ran a $2B+ marketplace across 130+ countries
Global research that scales — honestly
The research operating system for every market
EvidenceField
Every claim on this page is cited in the business plan.
EvidenceField0%