SYNTHETIC RESPONDENTS

Lewsearch

Runs synthetic focus groups with persistent respondents across a sequence of questions and follow-ups.

PlatformSynthetic respondents All providers Suggest an update

Focus-group workflow

Lewsearch describes discussions with five to 30 simulated participants across up to ten questions and follow-ups. Teams can introduce a PDF, image or URL, probe individual answers and review quotations, themes and a synthesis. The company positions these discussions as preparation for human interviews and concept tests. Scope and pricing are agreed with its team before work begins.

Lewsearch: synthetic focus groups

Respondent pipeline

Lewsearch says respondents start from American Community Survey census microdata. SwarmgramCity and the LewBrain substrate provide ongoing news, conversations and memory. Its Lewis 1.0 model asks the research questions; automated weighting, routing and calibration process the answers into study results.

Lewsearch: methodology and benchmark calculations

Current retrospective benchmark

As checked on 20 October 2026, Lewsearch reports 3.72 percentage points mean absolute error per answer and 3.89 points median error on its selected 500-question benchmark. The questions span 50 states, with 200 simulated respondents per question, and compare answer shares with published surveys, election results and official statistics. Lewsearch describes this as a calibrated retrospective evaluation, not a held-out or cross-validated result. Its methodology says the earlier 1,521-question benchmark is deprecated. These scores describe the published question set, rather than expected accuracy for a new client study.

Lewsearch: current homepage Lewsearch: methodology and benchmark calculations

Forecast results

Separately, the Lewsearch Report publishes 2.95% mean absolute percentage error (MAPE) across three graded spending outcomes as of 19 October 2026. The two back-to-school forecasts are averaged within one outcome, then the three outcomes receive equal weight. The methodology reports 2.09% MAPE for its listed persistence references using the same weights. This forecast measure is distinct from the retrospective benchmark's percentage-point error and does not evaluate focus-group dialogue or generated quotations.

Lewsearch: current homepage Lewsearch: methodology and benchmark calculations

Sources

Lewsearch: synthetic focus groups · Company product description
Lewsearch: current homepage · Company-reported results
Lewsearch: methodology and benchmark calculations · Company-reported methods and evaluation · 2026-10-19

Listing reviewed 20 October 2026