Collect real questions

Start with questions from search data, support conversations and sales notes you are allowed to use. Profound’s prompt-design guide also recommends combining SEO inputs with questions from support, sales and users. Treat that as a vendor’s workflow recommendation, not independent proof of demand. Source: Profound: prompt design.

Balance the sample

For a CRM study, include questions about finding a tool, comparing alternatives and checking a constraint such as email integration. Separate prompts naming your brand from prompts that leave the choice open.

A useful starter set can be small. Prefer a question with a clear reason for inclusion over many near-duplicates. Record the audience, intent and source of the question so another researcher can challenge your choices.

• Discovery: Which CRM tools suit a small consulting team?
• Comparison: How should a small team compare two CRM options?
• Constraint: Which CRM tools work with our existing email provider?
• Brand check: Does the named product support the required workflow?

Keep a stable baseline

Preserve exact wording for the core set. If you add a constraint or change the language, treat it as a new question or a new version. Otherwise, a shift in the answer may come from the changed prompt itself.

Keep exploratory questions separate from the baseline. They are useful for discovering topics, but adding them should not silently change the denominator in a trend report.

Check what already exists

Search the PromptFace library before defining a new monitoring scope. Search finds recorded questions; it does not generate a fresh answer. Read the prompt and answer date to see whether the record fits your study.

If a question is missing, use the suggestion form. Suggestions go to editorial review and do not start a measurement immediately. For a report, agree the question set, products, available dates and export fields before analysis begins.

• Plan content from your selected questions