# Evidence Check Reference

Use this file only after the user asks for research or agrees to an evidence check. Keep the result short and return to the same verdict framework.

## 30-60 Minute Evidence Check

Check in this order:

1. Trend signals: search trend, related queries, whether demand is rising, flat, or declining.
2. Complaint mining: reviews and community posts about existing tools, especially 1-4 star reviews.
3. Alternative search: what users compare, switch from, or use instead.
4. Payment signals: pricing pages, paid plans, paid complaints, preorder, budget, or agency/service spending clues.
5. Distribution clues: where the first users gather and whether manual outreach is realistic.

Do not treat search results as proof. Treat them as confidence signals.

## Useful Search Patterns

- `how to [solve problem]`
- `[competitor] overpriced`
- `[competitor] sucks`
- `[competitor] alternative`
- `[competitor A] vs [competitor B]`
- `better than [competitor]`
- `switched from [competitor] to`
- `best [tool category]`
- `[tool category] pricing`
- `[tool category] for [specific user]`
- `[specific user] [problem]`
- `[specific workflow] template`
- `[specific workflow] agency`

## Evidence Strength

Weak signals:

- Likes, compliments, and "I would use this"
- Friends saying it sounds cool
- Generic waitlists from non-target users
- Market size numbers without behavior

Medium signals:

- Search trends
- Repeated complaints about current tools
- Specific user interviews
- Painful current workarounds
- Multiple people asking for the same result

Strong signals:

- Users send real files, data, or workflows
- Users book calls or ask for access
- Users pay, preorder, or agree to pay
- Users repeatedly use a manual version
- Users switch from an existing solution

## No-Code Validation Actions

- Find 10 target users and ask how they solve the problem today.
- Search for complaints about current tools.
- Suggest a landing-page test with one clear promise and one real CTA; do not build it unless the user explicitly asks later.
- Manually deliver the result for 3 users.
- Ask users for real files, real data, or a real workflow.
- Try to collect a preorder, paid call, deposit, or booked demo.
- Compare current alternatives and identify the exact switching reason.

## Output Format

```markdown
## Evidence Check

Strongest positive signal:
- [What supports the idea.]

Strongest negative signal:
- [What weakens the idea.]

Biggest uncertainty:
- [What still must be validated directly with users.]

Verdict impact:
- [Keep / upgrade / downgrade the verdict.]

Next action:
- [One concrete action before building.]
```