Copilot Weekly Recap is the Nanga module that automatically generates a weekly performance brief for each market. Not a dashboard to interpret: a structured diagnosis, written by AI, that tells you what happened, why, and what to do next.
What you receive each week
For each dashboard (country or market group), Weekly Recap produces:
- A headline and summary — The week's trend in one sentence, centered on your primary KPI
- Key metrics — The 4 most meaningful indicators of the week, automatically qualified (positive, negative, neutral)
- Campaigns of interest — Up to 15 campaigns flagged as opportunities or problems, with diagnosis and recommendation
- A cross-market recap — A global synthesis sent via email or Slack, for a bird's-eye view in seconds
Comparison against a baseline, not week over week
The primary comparison is not "this week vs last week" — that is too volatile. Copilot Weekly Recap compares the current week to a 10-week historical baseline.
This approach smooths out natural variations (holidays, one-time spikes, seasonal dips) and separates real trends from noise. If your CAC is up 15% versus last week but sits within your historical corridor, the recap will not raise the alarm.
Week-over-week comparison remains available as secondary context.
Each campaign judged on the right KPI
One of the most important aspects of Copilot Weekly Recap: each campaign is evaluated on its own objective KPI, not a single metric applied across the entire portfolio.
How the KPI is determined
The system automatically identifies each campaign's objective by cross-referencing multiple information sources:
- Campaign name — If the naming convention indicates the objective (e.g., "appinstall", "registration"), the corresponding KPI is assigned automatically
- Platform-declared objective — Meta, TikTok, Google, Apple: each platform exposes the objective configured at campaign creation
- Client default KPI — If no signal is available, the account's primary KPI is used
An app acquisition campaign is evaluated on its CPI. An e-commerce conversion campaign on its CAC. A traffic campaign on its CPC. All without manual configuration — the mapping is intelligent and covers major platforms and campaign types.
Individual baselines per campaign
Campaigns are not compared to each other. Each campaign is compared to its own history.
Copilot Weekly Recap computes individual percentiles over each campaign's last 10 weeks. When the AI surfaces a metric, the system automatically qualifies whether that value falls within the campaign's normal corridor or breaks out of it — upward or downward.
Result: a high-CPM branding campaign will not be flagged as a "problem" just because its CPM is higher than a performance campaign's. It will only be flagged if its CPM exceeds its own historical norm.
Integrated business context
Copilot Weekly Recap does not stop at raw metrics. When available, the analysis integrates:
- Recent account changes — Budget modifications, targeting updates, campaign status changes. If your CAC rises after a budget doubling, the recap connects the dots.
- Business context — Information accumulated through conversations with Copilot (seasonality, product launches, market events).
This context enables the AI to produce more nuanced analyses that distinguish a real problem from an expected shift.
Structured AI generation
The recap is generated in a single call to Claude (Opus), Anthropic's most powerful model. The choice of Opus is deliberate: multi-campaign analysis requires cross-referencing many dimensions simultaneously — different objectives, different markets, different scales.
The AI receives aggregated and per-campaign data, analyzes meaningful variations, identifies probable causes, and produces a structured recap in a factual tone suited for a marketing director. No alarmist language, no superlatives: facts, context, recommendations.
Automatic metric qualification
Every metric surfaced in the recap is automatically qualified:
- Positive — The value is significantly better than the campaign's history
- Negative — The value is significantly below
- Neutral — The value sits within the normal corridor, or it is a volume metric (like budget) that carries no performance judgment
The "better/worse" direction adapts to the metric's nature: a rising ROAS is positive, a rising CPA is negative. Total cost is always neutral — it is an advertiser's choice, not a performance indicator.
Caching and reliability
Recaps are cached: requesting the same brief twice returns the same result. If a new market is added to an existing recap, only the missing dashboard is regenerated — the others are served instantly.
Summary
Copilot Weekly Recap transforms your advertising data into a structured, actionable weekly brief. Every campaign is evaluated on its own KPI, compared to its own baseline, and metrics are automatically qualified. All AI-generated, with zero manual configuration and integrated business context.