Key takeaways
- Write down the question before filtering data.
- Confirm metric definitions and timezones.
- Require more than one signal family for high-impact action.
- Schedule expiry and review for exclusions.
Before analysis
Capture the account, campaign, channel, date range, timezone, currency, landing pages, and conversion definitions. Note changes to budgets, bids, targeting, creatives, tags, consent, redirects, and the site.
- Check provider invalid-click columns and adjustments.
- Confirm tags loaded on every affected landing page.
- Test whether click identifiers survive redirects.
- Label incomplete or delayed evidence.
During investigation
Build a suspicious cohort and an appropriate comparison group. Review search terms, placements, locations, devices, schedules, network context, visitor continuity, behaviour, and outcomes. Keep raw facts separate from calculated signals and analyst notes.
- Check sample size before interpreting percentages.
- Group correlated signals to prevent double counting.
- Record alternative explanations.
- Use unclassified when confidence is insufficient.
After the decision
Record approval, action scope, provider response, verification, expected effect, monitoring period, expiry, and reversal. Compare later quality and legitimate volume, and retain provider-confirmed credits separately from internal estimates.
Limitations
What this guide does not claim
A checklist improves process consistency; it does not make incomplete data complete or turn suspicion into a provider-confirmed result.
Evidence
Primary sources
- Invalid clicks: definitionGoogle Ads Help
- About invalid trafficGoogle Ads Help
- Clicks and sessions discrepancy troubleshootingGoogle Ads Help
Read how we source, review, update, and correct content in our editorial standards.