# Case Study Template

Every published assay uses this structure. Numbers come from `summary.json` in the run record, never from memory.

---

## Assay {NNN}: {Target name}

**Protocol:** v{x.y} · **Date:** {YYYY-MM-DD} · **Run record:** {link}

### Headline

One sentence with the result, written like a lab finding.
Format: "{Variants} variants, {hits} hits, {gain}% over baseline on {primary function}, in {time}."

### Numbers

| Metric | Value |
|---|---|
| Variants generated | |
| Generations | |
| Hits | |
| Hit rate | |
| Baseline composite | |
| Best composite | |
| Gain over baseline | |
| Wall-clock time | |
| Compute cost (EUR) | |
| Human review time | |

### The target
What was designed, for whom, and why it mattered. Two to four sentences.

### The genome
Which genes were open, which were locked, and why.

### What the assay found
- The top finalist and what made it win, per function.
- One surprise: something the assay found that a human team would likely have skipped.
- One failure: a function or judge that misled, and how it was corrected.

### Human review
How many finalists were reviewed, accepted and rejected, with the most common rejection reasons.

### Field result (if run)
Metric, sample size, result, confidence.

### What changed in the protocol
Any new or revised fitness function this run produced.

### Before and after
Baseline vs winner, same viewport, side by side.

---

## Publishing rules

- Client work requires written approval before publishing. Anonymize when needed and say so.
- Never round numbers up. Report the exact values from the run record.
- If the run found no hits, publish it anyway with what was learned.