How InstaSights produces your results
Our partner Simsurveys develops machine learning models trained on millions of survey responses. InstaSights licenses these models to model consumer attitudes, preferences, and purchase intentions for your audience. Use the findings to evaluate ideas, names, messages, and prices.
Model provider: Simsurveys’ description of its training data and models.
How a study works
250 model-generated responses · Up to 10 questions
-
01
Your study
Your brief, audience, and questions give the model the context for your study. You review the setup before running it.
Product contextAudienceQuestions and answer choices -
02
Modeled consumer behavior
The Simsurveys model uses patterns learned from survey responses to model consumer attitudes, preferences, and purchase intentions for your product and audience. It generates individual responses to your questions without recruiting new participants.
Individual response data -
03
Your results
The answers are analyzed into a summary, results for each question, and comparisons between audience groups.
- Executive summary
- Audience comparisons
- Excel and CSV downloads
TRAINING DATA
Based on real consumer survey responses
Simsurveys describes its domain-specific models as trained on millions of validated survey responses to learn patterns in consumer attitudes, preferences, and purchase intentions. Your product context, audience, and questionnaire guide the predictions for your study.
Use the results to compare concepts, narrow a list of names, refine a message, or understand reactions to a price.
EVIDENCE AND INTERPRETATION
How to judge the findings
Published evaluations compare modeled responses with answers collected from people. Check how closely the questions and audience match your study, and look at the differences in the answers.
See a published comparison
Importance of free returns when choosing an online retailer
| Human survey | Machine learning model |
|---|---|
| 76% | 88% |
The model’s estimate is 12 percentage points higher. This example compares one question from a retail-returns study. Accuracy can vary by question and audience.
2,007 human respondents and 2,007 model-generated responses. Source: a model-provider comparison with the Happy Returns / NRF 2024 survey.
Read the source report (PDF)Use the findings to make informed decisions
Compare the full range of answers and the size of each audience group, especially when results are close. The findings describe consumer attitudes and likely choices; sales, product use, and campaign performance are measured through actual activity.
Read the evidence checklistReading the examples
Know what each example shows
The oat-latte naming study contains model-generated responses. The October 15, 2026 final study includes all 250 responses, question results and audience comparisons in its downloads. It demonstrates results you can inspect; it is not a comparison with a human naming survey.
The water-bottle examples on the product page and five solution pages are actual Simsurveys simConsumer model outputs, recorded October 15, 2026. Each is an independent single-choice question without demographic filters or conditioning. Percentages are predicted probabilities, not respondent counts; these examples are not complete studies. Download the exact questions, context and results.
The retail-returns comparison above is a separate model-provider evaluation against a human survey. Its 12-percentage-point difference applies to the stated question, not an overall accuracy score. Use the evidence worksheet to record what is supported and what still needs testing for your decision.
Start your first study
One complete study for $99
Plan your next study
- Compare synthetic research tools by workflow and evidenceA documentation-based shortlist of InstaSights, Synthetic Users and Qualtrics Synthetic Panels, with questions about scope, outputs, cost and validation.