Morning Brief

Microsoft Ads Morning Brief

Produce a daily, prioritized account-health narrative for Microsoft Advertising campaigns with actionable items, budget pacing vs monthly target projection, bot traffic monitoring, and import drift detection.

Account Context

Read ${CLAUDE_PLUGIN_ROOT}/profile/account-profile.md at the start of every run. If it exists:

• Use known account IDs -- skip list_accounts discovery.
• Apply KPI targets as anomaly detection thresholds (e.g., flag CPA > target CPA).
• Note active tests when interpreting performance shifts.
• Check watch list for follow-up items from prior sessions.
• Read Import Config for auto-import drift detection. If it doesn't exist, fall back to list_accounts and suggest running platform-setup.

Data Access

• mcp__bing-ads__report: Generate performance reports (campaign, keyword, search query). Returns parsed CSV data as JSON.
• mcp__bing-ads__query: Query campaign structure (campaigns, ad groups, keywords, ads). Read-only.
• mcp__bing-ads__list_accounts: Validate account access and discover account IDs.

Use the report tool with report_type and date_range parameters. See references/bing-queries.md for report configurations.

Connection verification

1. If the profile has account IDs, use them directly.
2. Otherwise, run mcp__bing-ads__list_accounts to discover accounts.
3. If it fails, report the connection failure and suggest running platform-setup.
4. If auth fails with token error, note that Microsoft rotates refresh tokens and suggest re-running the OAuth flow.

Workflow

Phase 1: Collect data

Execute four data pulls from references/bing-queries.md:

1. MB-1: Campaign performance report (30d, Daily aggregation).
2. MB-2: Keyword performance report (Yesterday, Daily aggregation).
3. MB-3: Campaign structure query for budget, status, and network settings.
4. MB-4: Search query report (Yesterday) for bot traffic signal detection.

Run reports 1, 2, and 4 in parallel. Query 3 depends on account access confirmation.

Phase 2: Detect and rank anomalies

Use references/anomaly-formulas.md for exact formulas. The formulas are platform-agnostic -- they operate on metric values, not platform-specific fields.

For each campaign, for each metric (cost, conversions, CPA, CTR):

1. Compute baseline_7d = mean of last 7 days (excluding yesterday).
2. Compute baseline_30d = mean of last 30 days (excluding yesterday).
3. Compute deviation_pct = (yesterday - baseline_7d) / baseline_7d.
4. Compute dollar_impact = yesterday_value - baseline_7d (for cost/CPA metrics). CTR is not dollar-denominated; use deviation_pct only.
5. Surface when BOTH: |deviation_pct| > 0.20 AND |dollar_impact| > $10. For CTR: surface when |deviation_pct| > 0.25 (no dollar gate).
6. Rank all flagged items by |dollar_impact| descending.

Cap output at 10 anomaly items to keep the brief actionable.

Phase 3: Bot traffic monitoring

Analyze yesterday's data for suspicious click patterns that indicate bot or invalid traffic:

1. High click, zero conversion campaigns: Flag campaigns with clicks > 50 and conversions = 0 yesterday.
2. Geographic anomalies: If geo data is available, flag unexpected spikes from geos not in targeting.
3. Device anomalies: Flag campaigns where a single device type accounts for >80% of clicks with zero conversions.
4. CTR outliers: Flag keywords or campaigns with CTR > 15% combined with zero conversions (likely click fraud).

Present bot traffic signals as a separate section. These are signals, not confirmations -- recommend checking Microsoft Advertising's Invalid Clicks report in the UI.

Phase 4: Import drift detection

If the account profile has Import Config data:

1. Check if auto-import is enabled and the last known import date.
2. Compare current campaign settings (from MB-3) against expected settings:
   • MSAN distribution unexpectedly enabled (common auto-import side effect).
   • Search partners unexpectedly enabled.
   • Budget changes that don't match manual adjustments in the Decision Log.
3. Flag any drift as a Watch item with the specific setting that changed.

If no Import Config exists, skip this phase.

Phase 5: Budget pacing assessment

Bing campaign budgets are retrieved via mcp__bing-ads__query (campaigns entity returns daily_budget).

For each campaign:

1. daily_budget = campaign daily budget from structure query.
2. monthly_budget = daily_budget * 30.4.
3. day_of_month = calendar day number.
4. expected_mtd_spend = daily_budget * day_of_month.
5. actual_mtd_spend = sum of daily cost for current month from MB-1.
6. pacing_ratio = actual_mtd_spend / expected_mtd_spend.
7. Flag overpacing when pacing_ratio > 1.10 and underpacing when pacing_ratio < 0.85.

Monthly target projection

If the account profile includes a Monthly Budget target:

1. total_mtd_spend = sum of all campaign MTD spend.
2. projected_eom_spend = total_mtd_spend / day_of_month * days_in_month.
3. budget_utilization = projected_eom_spend / monthly_budget_target.
4. Flag when projected spend exceeds target by >10% or falls below target by >15%.

Present both campaign-level pacing and account-level monthly projection.

Phase 6: Draft prioritized narrative

Structure output with three priority tiers:

• Urgent: needs action today.
• Watch: monitor or schedule action.
• Healthy: stable areas.

Every item must include the likely cause and one concrete next action with a Microsoft Advertising UI path.

Output format

## Morning Brief - [Date]

### Account
- Microsoft Ads: [Account Name] ([Account ID]) - [N] campaigns active

**Overall:** [one-sentence account summary]

### Budget Pacing
| Campaign | Daily Budget | MTD Spend | Expected MTD | Pacing | Status |
|---|---:|---:|---:|---:|---|

**Monthly Projection:** $X,XXX projected vs $X,XXX target ([over/under] by X%)

### Bot Traffic Signals
| Campaign | Signal | Clicks | Conv | CTR | Recommended Check |
|---|---|---:|---:|---:|---|

(If no signals detected: "No suspicious click patterns detected yesterday.")

### Import Drift
- [drift findings or "No import drift detected."]

### Urgent
1. [Issue + dollar impact + recommended action + UI path]

### Watch
1. [Issue + dollar impact + recommended action + UI path]

### Healthy
- [stability observation]

### Notes
- Data freshness: Bing reports reflect data through yesterday's close.
- **Limitations**: No change event history available via API. No ad disapproval data in reports -- check Microsoft Advertising UI > Ad Extensions & Ads > Ads for editorial status. No impression share in standard reports -- use budget pacing as a proxy for coverage.

Guardrails

• Conversion lag: When yesterday conversions are >30% below 7d baseline, add a note that conversions typically backfill for 24-72 hours. Do not flag as "Urgent" unless the drop also appears in 2-day-old data.
• Bing data differences: Spend is already in dollars (no micros conversion). CTR is returned as a percentage string -- parse to float for math.
• No change events: Bing API does not expose account change logs. Note this limitation rather than omitting historical context.
• No disapproval data: Ad editorial status is not available via the reporting API. Note this in the output.
• No impression share: Standard Bing reports do not include impression share. Use budget pacing and spend-to-budget ratio as proxy signals for coverage gaps.
• Distinguish between "no issues found" and "insufficient data".
• Keep recommendations operational and specific with UI paths from references/ui-paths.md.
• Cap each priority section: max 5 Urgent, 5 Watch, 5 Healthy items.
• Read-only: This skill produces analysis and recommendations only. No account modifications are made.

Profile Maintenance

After completing analysis, if ${CLAUDE_PLUGIN_ROOT}/profile/account-profile.md exists:

1. Update Watch List with any new anomalies flagged in this run.
2. Update Active Tests if user mentioned starting or completing a test.
3. Append to Decision Log if actions were taken (pauses, negatives added, etc.).
4. Update "Last updated" date. Present proposed profile changes to the user before writing.

References

• references/bing-queries.md
• references/anomaly-formulas.md
• references/ui-paths.md