What Silicon Samples Are For
Silicon Samples generate AI-created test responses so you can inspect survey logic, export structure, cleaning workflows, and analysis pipelines before collecting real participant data. They are useful for dry runs, demos, and stress-testing branching designs.
Silicon Samples are not a substitute for human pilot testing. Treat generated responses as synthetic data: useful for workflow checks, but not evidence about real respondents.
Export: Silicon Samples
Choose A Sample Source
Step 1: Open Export. Silicon Samples live with response and export workflows because generated responses appear alongside other survey responses.
Step 1: Open Export
Step 2: Choose Describe, Past Survey, or Upload CSV. Describe is best for broad personas, Past Survey is best for matching earlier respondent patterns, and CSV is best when you already have a structured sample frame.
Step 3: Review the persona fields. Make sure generated respondents represent the test cases you need, such as screen-outs, treatment arms, edge cases, and missing-data scenarios.
Step 4: Select an AI credential. Use a saved provider key approved for the workspace and study context.
Step 4: Saved Credentials
Step 5: Generate the sample. Wait for the job to finish before exporting or cleaning the generated responses.
Review Generated Responses
- Check routing coverage. Confirm generated respondents reached the expected branches, endings, and quota-relevant paths.
- Check variables. Look for missing question names, unclear recodes, or columns that would confuse analysis.
- Check response realism only as a smoke test. Do not infer population behavior from generated answers.
- Check downstream tools. Use the generated rows to test CSV/XLSX downloads, API pulls, codebooks, scripts, and dashboards.
Export: Generated Data
Clean Up Before Real Collection
Generated responses are tagged so you can identify them in Export. Delete or exclude them before analysis, and make sure the live dataset does not mix synthetic rows with participant responses.
- Export a test copy if useful. Keep a local dry-run export only if it helps document your pipeline test.
- Delete generated rows when done. Remove Silicon Sample responses before live collection or final analysis.
- Run one human pilot afterward. A human pilot can catch wording, comprehension, and device issues that synthetic responses cannot.
Preview: Human Pilot