---
name: coppica-signal-report
description: Use when the user wants to know what's working for a client, what to scale, and what to kill. Turns Coppica's attribution and performance data into a plain-language recommendation.
---

# Signal report

Translate Coppica's attribution data into a recommendation a human can act on. This is the
read-side counterpart to the optimization loop.

## Steps

1. **Pull the brief.** Generate or fetch the performance brief and the attribution summary for the
   client (the attribution endpoints expose the canonical conversion ledger, a per-output summary,
   and a generated brief).
2. **Rank for context.** Use `coppica_rank_outputs` to see how current pieces order by signal, and
   note the `confidence_tier` so you can caveat appropriately.
3. **Diagnose the notable ones.** `coppica_diagnose_output` on the standout winners and losers to
   explain *why*, in technique terms, not just *that* they performed.
4. **Write the recommendation.** In plain language: what is working and should be scaled, what is
   underperforming and should be paused or reworked, and what to test next. Always state the
   confidence level honestly; this is signal-informed prioritization, not prediction.

## Decision branches

- **Thin data / low confidence?** Say so plainly. Recommend keeping copy running and reporting
  outcomes rather than over-reading early numbers.
- **User wants action, not just a report?** Hand off to `coppica-optimization-loop` to actually
  iterate on the winners you surfaced.

## Verify

Your recommendation cites specific outputs and their outcomes, and states a confidence level.

## Next

`coppica-optimization-loop` to act on it.