Waste Detector

Paid Search Waste Detector

Scan Google Ads and Bing Ads paid search accounts for the most common spend leaks and quantify each leak in dollars with an action plan.

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. If it doesn't exist, fall back to list_accounts and suggest running platform-setup.

Data Access

Google Ads

• mcp__google-ads__query: Execute GAQL SELECT queries for waste detection.
• mcp__google-ads__mutate: Stage and apply mutation operations.
• mcp__google-ads__list_accounts: Validate account visibility before running audits.

Bing Ads

• mcp__bing-ads__report: Generate performance reports for waste analysis.
• mcp__bing-ads__query: Query campaign structure (campaigns, ad groups, keywords, ads).
• mcp__bing-ads__list_accounts: Validate account access.

Platform detection

Follow the same pattern as morning-brief: try both platforms, run whichever responds, gracefully skip the other.

Mutation safety flow (Google only — Bing MCP does not yet support mutations)

1. Always run mcp__google-ads__mutate with dry_run: true first.
2. Show exactly what would change.
3. Require explicit user approval.
4. Re-run with dry_run: false only after approval.

For Bing waste findings, present recommendations as manual action items (no automated mutations available yet).

Workflow

Phase 1: Run waste queries

Google Ads

Execute the eight query groups from references/gaql-queries.md.

Bing Ads

Execute the equivalent Bing queries from references/bing-queries.md.

Waste types covered (both platforms where data is available):

Phase 2: Quantify impact

Use references/thresholds.md for exact detection thresholds and dollar formulas per waste type. Summary:

1. Apply the specific threshold for each waste type (not a single blanket rule).
2. Compute dollar waste using the formula specified per type.
3. Rank all findings by dollar impact descending, across both platforms.
4. Tag severity: HIGH (>$500/mo), MEDIUM ($100-500), LOW ($25-100), INFO (<$25).

Use references/benchmarks.md for QS-to-CPC pressure multipliers and CTR bands. These benchmarks apply to both Google and Bing (industry-standard metrics).

Phase 3: Build remediation package

Map each finding to its remediation action using references/thresholds.md remediation mapping.

Google Ads (automated mutations available)

• Types 1, 2, 7: build pause operations and preview with mcp__google-ads__mutate dry run.
• Type 8: build negative-keyword operations and preview with mcp__google-ads__mutate dry run.
• Types 3, 4, 5: recommend manual UI changes (cannot be mutated via API).
• Type 6: recommend creating additional RSA variants.

All mutations must follow dry-run first semantics. Show preview, get user approval.

Bing Ads (manual recommendations only)

All Bing waste findings are presented as actionable recommendations with specific steps to take in the Microsoft Advertising UI. Format each as: entity name, problem, dollar impact, and the exact UI path to fix it.

Output format

## Waste Report - [Date]

### Google Ads: [Name] ([Customer ID])
**Estimated Recoverable Waste: $X,XXX/month**

| # | Waste Type | Monthly Cost | Severity | Action |
|---|---|---:|---|---|

### Bing Ads: [Name] ([Account ID])
**Estimated Recoverable Waste: $X,XXX/month**

| # | Waste Type | Monthly Cost | Severity | Action |
|---|---|---:|---|---|

### Cross-Platform Summary
**Total Recoverable Waste: $X,XXX/month** (Google: $X,XXX + Bing: $X,XXX)

### Detailed Findings
- **[Google/Bing]** [Entity], [problem], [dollar impact], [recommended action]

### Ready-to-Apply Changes (Google Ads)
- [mutation preview]

### Manual Action Items (Bing Ads)
- [recommendation with UI path]

Guardrails

• Call out where dollar figures are estimates vs direct spend totals.
• Keep assumptions explicit for model-based calculations.
• Do not run live mutations without explicit user approval.
• Distinguish "true zero" from omitted zero-value GAQL rows.
• Bing limitations: Note that Bing waste findings cannot be auto-remediated. Present clear manual steps.

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/gaql-queries.md
• references/bing-queries.md
• references/thresholds.md
• references/benchmarks.md
• references/google-campaign-management.md