Answers

Synthetic respondents accuracy

Synthetic-respondent accuracy is not one number. Vendors publish mean absolute error, Spearman correlation, or thematic parity. Those units do not convert. Lewsearch publishes 7.47% MAE on 404 ex-electoral questions. Ask every other vendor for the instrument, the n, and the unit before you compare.

Three units on three homepages

Lewsearch scores option shares against public polls. Methodology: /methodology. Coverage: /coverage. Synthetic Users, on syntheticusers.com, reports 85–92% synthetic-organic parity on qualitative interviews. Aaru, on aaru.com/simulation, reports 0.90 median Spearman on a blinded EY wealth study (3,600 investors, 30+ markets). Simile, on simile.ai, reports over 7,000 weekly evaluations and does not publish a pooled MAE.

Worked example: why 85% is not 7.47%

Suppose a human interview and a synthetic interview both mention price, trust, and packaging. Thematic overlap can be high even if the synthetic panel would have put "would buy" at 62% and the real survey put it at 41%. That 21-point gap is MAE territory. The reverse is also true: option shares can be close while the quotes feel generic. If you buy interviews, use parity. If you buy a crosstab, use MAE. If you buy a rank-order of wealth attitudes, a Spearman on that study is the relevant published cell.

Accuracy units in synthetic research, as published
Vendor (own site)UnitWhat it measures
LewsearchMAE 7.47%Option-share gap vs public polls
AaruSpearman 0.90 (EY study)Rank agreement on that recreation
Synthetic Users85–92% parityTheme overlap on interviews
Simile7,000+ weekly evalsValidation volume; no pooled MAE

Sources

FAQ

How accurate are synthetic respondents?
It depends on the unit. Lewsearch publishes 7.47% MAE on 404 public-poll items. Synthetic Users publishes 85–92% thematic parity on qualitative interviews. Aaru publishes a 0.90 Spearman on one EY wealth study. Do not average those figures.
What is MAE?
Mean absolute error: the average absolute difference, in percentage points, between each predicted option share and the published survey share.
What is thematic parity?
Synthetic Users defines it on its homepage as overlap in themes, depth, and qualitative alignment between synthetic and organic interviews. It is not option-share error.
What is Spearman correlation here?
Aaru reports a median Spearman of 0.90 on a blinded EY Global Wealth Study recreation. Spearman measures rank agreement. It is not MAE.

Check the work

Related

Canonical: https://lewsearch.com/answers/synthetic-respondents-accuracy