skill-019-keyword-cannibalization-analysis

Skill 019: Keyword Cannibalization Analysis

Purpose

Detect when multiple keywords within the account compete for the same search queries, leading to internal auction competition and suboptimal ad serving. Cannibalization splits impressions and routes traffic to lower-performing keywords.

Data Requirements

Data Source: Standard

Standard Data:

• data/performance/campaigns/*/*/search_terms_metrics_30_days.md - Search query data with matched keyword
• data/account/campaigns/*/*/keywords.md - Keyword settings and match types
• data/performance/campaigns/*/*/keywords_metrics_30_days.md - Keyword performance

Reference GAQL:

SELECT
  campaign.name,
  ad_group.name,
  search_term_view.search_term,
  segments.keyword.info.text,
  metrics.impressions,
  metrics.clicks,
  metrics.conversions
FROM search_term_view
WHERE segments.date DURING LAST_30_DAYS
  AND metrics.impressions > 0

Use /google-ads:get-custom if you need different date ranges.

Analysis Steps

1. Map search terms to keywords: For each search term, identify which keyword/ad group triggered it
2. Identify multi-trigger terms: Find search terms appearing in multiple ad groups or campaigns
3. Analyze keyword overlap: Compare match types, Quality Scores, and performance of cannibalizing keywords
4. Calculate impact: Total impressions split, potential CPA improvement if consolidated
5. Determine root cause: Overlapping match types, similar keywords without negatives, poor ad group structure

Thresholds

Condition	Severity
Search term appears in 3+ ad groups	Critical
Low-QS keyword winning over high-QS	Critical
High-CPA keyword getting 50%+ impressions	Warning
Broad match stealing from exact match	Warning
Cross-campaign overlap	Info

Output

Use Short format by default. Use Detailed if user requests comprehensive analysis.

Short:

## Cannibalization Analysis Audit
**Account:** [Name] | **Overlapping Terms:** [X] | **Issues:** [Y]

### Critical ([Count])
- **"[search term]"**: Triggers [X] keywords, low-QS variant winning → Add cross-negatives
- **[Keyword A] vs [Keyword B]**: [X]% overlap, wrong keyword winning → Route to best performer

### Warnings ([Count])
- **Broad match cannibalization**: "[broad]" stealing from "[exact]" → Add negative to broad ad group

### Recommendations
1. Add cross-negatives for [X] keyword pairs
2. Consolidate overlapping ad groups