skills/monday-briefing/examples/output.md

> Reference output. Numbers from `evals/fixtures/ecommerce-healthy.mjs` (7d).
> Hard rules this example demonstrates: under 30 lines, one opportunity only,
> no statistical caveats inside the block, and no "good job" framing.

```
🦭 Seal Copilot — Monday Briefing · demo-store.com · 30 Aug – 5 Sep

🎯 VERDICT: ✅ on track
Conversions and revenue grew slightly faster than traffic; nothing needs action.

📊 THIS WEEK vs LAST
| Entrances | CR | Conversions | Revenue | AOV |
| 9,850 (+2.5%) | 2.35% | 231 (+3.1%) | €17,900 (+3.8%) | €77.49 |

🏆 WHAT WORKED
brand-es: 34 conversions from 930 entrances (3.66% CR), the best line in paid.

🩹 WHAT NEEDS ATTENTION
retarget-es: 645 entrances at 1.09% CR, half the channel average. Third week.

💰 TOP OPPORTUNITY THIS WEEK
Leaky campaign — retarget-es takes 28% of Paid Search traffic and returns 12%
of its conversions → audit the audience and creative before renewing spend →
≈€480/month if it reaches the 2.39% channel average → re-check in 2 weeks.

🚨 WATCHDOG
Add-to-cart: 🟢 normal, tracking against the 4-week baseline.
Tracking decay (microconversions): 🟢
Bot share: 6% (last week 6%)

➡️ NEXT
Suggested follow-up: "Diagnose retarget-es."
```

Notes for the model, not part of the output: the follow-up block is omitted
entirely when nothing was due for verification. When something was due, it goes
directly above 🚨 WATCHDOG as one line per recommendation checked.