Compare
Lewsearch vs Simile
Lewsearch publishes a calibrated result. On the calibrated 50-state benchmark, the average miss is 7.07 points on the best 80% of 1,521 scored questions, and 10.02 points across all of them. Simile publishes 7,000+ evaluations on a weekly-validated behavior model and does not publish a percentage-point MAE (simile.com, checked October 2, 2026). A Simile post announces a $100 million round with Index (simile.ai/blog/the-simulation-company, checked October 2, 2026). Studies on Lewsearch are set up with the team. Pricing is scoped to your audiences and study volume on a short call. Your first qualified U.S. consumer Blind Challenge is free. Enterprise teams start with an 8-week pilot that rolls into an annual engagement.
What each site actually says
Simile's homepage describes a weekly-validated behavior model, with 7,000+ evaluations, and the call to action is “Run a simulation.” Named quotes include CVS, Wealthfront, Itaú, Suntory, Gallup, and Garnett Station. The homepage also cites a 1,000-person paper (simile.com, checked October 2, 2026).
Lewsearch publishes a public-poll miss. On the calibrated 50-state benchmark, the average miss is 7.07 points on the best 80% of 1,521 scored questions, and 10.02 points across all of them. Every study is calibrated against published survey data before it reaches you. Enterprise panels are also calibrated to your own past studies, so the numbers line up with the tracker you already trust. The scored file is on /benchmarks/states. Lewsearch keeps 650,000 census-grounded simulated Americans across 50 states and D.C..
Worked example: one category, one n
On the calibrated 50-state benchmark, the average miss is 7.07 points on the best 80% of 1,521 scored questions, and 10.02 points across all of them. Each question in that file uses 200 simulated respondents. Every study is calibrated against published survey data before it reaches you. Enterprise panels are also calibrated to your own past studies, so the numbers line up with the tracker you already trust. A narrower April test on 5 places with 10,000 respondents scored 7.47 points on 404 non-electoral questions.
Simile's 7,000+ evaluations are a volume claim. The 1,000-person paper is a separate academic citation on the homepage. This page does not treat either as a production MAE against public polls.
| Rubric | Lewsearch | Simile |
|---|---|---|
| Published error metric | Calibrated 50-state benchmark: 7.07 points on the best 80% of 1,521 scored questions, 10.02 across all of them. | 7,000+ evaluations. No percentage-point MAE (simile.com, checked October 2, 2026) |
| Named benchmarks | Gallup/Canvass, UT/Texas Politics Project, PPIC. Public file of 1,521 scored questions. | Weekly evaluations. A 1,000-person paper is cited on the homepage (simile.com, checked October 2, 2026). |
| n disclosed | 1,521 scored questions. 200 simulated respondents per question. | 7,000+ evaluations (simile.com, checked October 2, 2026). |
| Public pricing | Scoped on a call. First qualified U.S. consumer Blind Challenge is free. Enterprise starts with an 8-week pilot. | No price on the homepage (simile.com, checked October 2, 2026). |
| Access model | Book a demo. Studies are delivered with the team. | Homepage CTA: Run a simulation (simile.com, checked October 2, 2026). |
| Company fact on their site | 650,000 census-grounded simulated Americans across 50 states and D.C.. Forecasts graded in public: 3 wins, 1 miss, 3 awaiting a grade. | $100 million round with Index (simile.ai/blog/the-simulation-company, checked October 2, 2026). |
Where each product fits
If the job is behavioral simulation of how people respond to change, Simile publishes that product (simile.com, checked October 2, 2026). Lewsearch returns respondent-level records, so crosstabs add up to the total, and the same simulated respondents can be re-asked across waves. Audiences can be fit to a client's own data and scored on the client's own past studies. Enterprise workloads run with no external model API calls. Book a demo. The team sets up your first study with you.
Sources
Competitor cells cite Simile's own pages, checked October 2, 2026. Lewsearch cells cite the public benchmark file.
FAQ
- Does Simile publish a percentage-point error rate?
- Simile's homepage publishes 7,000+ evaluations and does not publish a percentage-point MAE (simile.com, checked October 2, 2026).
- What accuracy number does Lewsearch publish?
- On the calibrated 50-state benchmark, the average miss is 7.07 points on the best 80% of 1,521 scored questions, and 10.02 points across all of them. A narrower April test on 5 places with 10,000 respondents scored 7.47 points on 404 non-electoral questions. Every study is calibrated against published survey data before it reaches you. Enterprise panels are also calibrated to your own past studies, so the numbers line up with the tracker you already trust. The public file is at /benchmarks/states.
- How much has Simile raised?
- A Simile post announces a $100 million round with Index (simile.ai/blog/the-simulation-company, checked October 2, 2026). That is a company fact. The product site is simile.com.
- Which one has public pricing?
- Pricing is scoped to your audiences and study volume on a short call. Your first qualified U.S. consumer Blind Challenge is free. Enterprise teams start with an 8-week pilot that rolls into an annual engagement. The pricing page explains engagement options. Simile does not print a price on its homepage (simile.com, checked October 2, 2026).
Check the work
Related comparisons
Canonical: https://lewsearch.com/compare/lewsearch-vs-simile