# Evidence Grading — Market Research

Use this ladder to prevent weak research from sounding stronger than it is.

## Confidence Ladder

| Confidence | Typical Evidence | How to Use It |
|------------|------------------|---------------|
| **High** | Public filings, first-party pricing pages, direct customer behavior, signed pilots, prepayments | Strong enough to support a recommendation |
| **Medium** | Review mining, job posts, trend data, credible analyst reports, repeated interviews | Useful for directional judgment with caveats |
| **Low** | Founder claims, press releases, one-off anecdotes, generic social chatter | Use only as a lead, never as the core conclusion |

## Minimum Standard by Decision

| Decision Type | Minimum Bar |
|---------------|-------------|
| Kill or continue an idea | Medium confidence from multiple source families |
| Pricing recommendation | Medium-to-high confidence plus direct customer evidence |
| Market entry recommendation | High confidence on demand, competition, and reachability |
| Expansion timing | High confidence on segment fit and local constraints |

## Output Rule

Every major conclusion should include:
- the claim
- the supporting evidence
- the confidence level
- what would invalidate it

If you cannot state all four clearly, the conclusion is not ready.