Google Ads Tracking

Read first: - google-ads/references/operator-thesis.md - google-ads/references/tracking-playbook.md - google-ads/references/deliverable-templates.md

Read workspace if available: - workspace/ads/account.md - workspace/ads/goals.md - workspace/ads/findings.md - workspace/ads/change-log.md - workspace/ads/learnings.md

Data Acquisition

Connected Mode (MCP available)

Pull via the search tool on google-ads-mcp:

Primary: Conversion actions and configuration:

SELECT
  conversion_action.id,
  conversion_action.name,
  conversion_action.type,
  conversion_action.category,
  conversion_action.status,
  conversion_action.counting_type,
  conversion_action.include_in_conversions_metric,
  conversion_action.value_settings.default_value
FROM conversion_action
WHERE conversion_action.status = 'ENABLED'

Primary: Conversion performance by action (detect duplicates/pollution):

SELECT
  conversion_action.name,
  conversion_action.type,
  conversion_action.category,
  metrics.conversions,
  metrics.conversions_value,
  metrics.all_conversions
FROM conversion_action
WHERE segments.date DURING LAST_30_DAYS
  AND conversion_action.status = 'ENABLED'
ORDER BY metrics.conversions DESC

Supplementary: Account auto-tagging status:

SELECT
  customer.id,
  customer.descriptive_name,
  customer.auto_tagging_enabled
FROM customer

Supplementary: Campaign-level conversion metrics (look for anomalies):

SELECT
  campaign.name,
  metrics.conversions,
  metrics.all_conversions,
  metrics.conversions_value,
  metrics.cost_micros,
  metrics.cost_per_conversion
FROM campaign
WHERE campaign.status = 'ENABLED'
  AND segments.date DURING LAST_30_DAYS
ORDER BY metrics.conversions DESC

See data/gaql-recipes.md for additional queries.

Date Range Fallback

Conversion action configuration queries (no date range) always work. For conversion performance queries, if LAST_30_DAYS returns zero data, fall back to LAST_90_DAYS, then all-time. Configuration queries don't need fallback — they reflect current state regardless. Always state the date range used.

Export Mode (no MCP)

Ask the user for: - Conversion actions list (name, type, counting, include in conversions, value settings) - Google Tag Assistant screenshot or diagnosis notes - GA4 conversion import settings (if applicable) - Any notes on tracking implementation

See data/export-formats.md for recommended format.

Diagnostic Framework

1. Can we trust the account?

Assign a tracking confidence level immediately using the rubric in google-ads/references/tracking-playbook.md.

Threshold checklist (from tracking-playbook.md): - Count primary conversion actions (should be 1-2 clear ones) - Check include_in_conversions_metric — are micro-conversions polluting primary? - Compare conversions vs all_conversions — if ratio is >2x, investigate - Check counting type — "One" for leads, "Every" for e-commerce only - Check for duplicate sources — same event tracked by both native tag AND GA4 import? - Check auto-tagging status - Verify conversion volume (≥15/30d for HIGH, 5-14 for MEDIUM, <5 for LOW)

2. What to check

Duplicate counting: - Same conversion tracked by both native tag AND GA4 import - Multiple conversion actions for the same event - "Every" counting type on lead/form events (should be "One")

Micro-conversion pollution: - Page views, scroll depth, time on site set as primary conversions - These inflate conversion count and deflate apparent CPA - Smart bidding optimizes toward them instead of real conversions

Missing conversions: - Key events not tracked at all - Phone calls without call tracking - Form submissions without confirmation page or event tracking

Attribution concerns: - Auto-tagging disabled (GCLID not passing) - Cross-domain tracking gaps - Conversion window too short or too long

Value quality: - All conversions have the same default value - Values don't reflect actual revenue - Missing value tracking on e-commerce

3. Impact assessment

For each problem found, assess: - How much is this distorting reported performance? - Which campaigns are most affected? - Is smart bidding learning from bad signal? - What decisions should wait until this is fixed?

Draft Output

Tracking Fix Draft

Trigger: Any tracking problem at Medium confidence or lower.

Create using drafts/templates/tracking-draft.md: - Write to workspace/ads/drafts/YYYY-MM-DD-[account-slug]-tracking-fix.md - Include specific fixes: what to change, current vs. proposed state - Note the impact on reported conversions ("Conversions will appear to drop ~40%") - Note the impact on smart bidding ("2-3 week recalibration period") - Critically: list which optimization decisions should wait until tracking is fixed - Update workspace/ads/drafts/_index.md

Blocking behavior

When tracking confidence is Low or Broken, the draft should include a "blocked decisions" section that explicitly names what other skills/drafts should NOT proceed until tracking is resolved: - Budget scaling → blocked - Structure changes → proceed with caution - Negatives → can proceed (waste is waste regardless of tracking) - RSA changes → can proceed (copy quality is independent)

Always update workspace memory:

• workspace/ads/findings.md — tracking diagnosis and confidence level
• workspace/ads/account.md — update tracking status if it has changed
• workspace/ads/learnings.md — what we learned about this account's tracking

Output Shape

1. Account Status block — account name, CID, status, date range used, tracking confidence, mode
2. Conversion actions inventory (what's tracked, how, counting type)
3. Problems identified (with severity and evidence)
4. Impact assessment (what's distorted, which campaigns affected)
5. Recommended fixes (specific changes)
6. Blocked decisions (what should wait)
7. Draft created (path and summary)
8. Memory updates

Rules

• If tracking is unreliable, say it early.
• Do not bury trust issues under optimization tips.
• Explain what decisions should wait until tracking is fixed.
• Tracking confidence is a gate. If it's Low or Broken, other skills should know.
• The most dangerous Google Ads accounts look great on paper because tracking is counting the wrong things. Surface this clearly.