Key takeaways
- Define clicks, sessions, visitors, and conversions separately.
- Every rate needs a visible denominator.
- Coverage and eligibility belong with the result.
- Provider and first-party metrics should be reconciled, not blended silently.
Choose one unit for each question
Use provider clicks for ad interaction questions, eligible measured sessions for on-site quality, visitors for continuity, conversion occurrences for outcomes, and provider credits for confirmed billing adjustments. Switching units mid-calculation makes a polished metric unreliable.
Document website, channel, timezone, currency, date range, attribution window, and source before comparison.
Publish the metric envelope
A result should include its numerator, denominator, eligibility rule, excluded population, coverage, freshness, confidence, and known limitations. If the denominator changes after late data arrives, retain the earlier state and reconciliation.
- Use unavailable for missing evidence.
- Use insufficient for samples below a decision threshold.
- Use unclassified for unresolved risk.
- Use stale when freshness exceeds the supported window.
Avoid the savings shortcut
Analyzed spend, spend associated with suspicious traffic, protected spend, estimated impact, and confirmed provider credit are different. A risk score is not a probability and should not be multiplied by cost as an automatic savings calculation.
Limitations
What this guide does not claim
Traffic quality is not one universal score. Comparability depends on compatible definitions, channel context, tracking coverage, and sufficient sample size.
Evidence
Primary sources
- Clicks and sessions discrepancy troubleshootingGoogle Ads Help
- About invalid trafficGoogle Ads Help
- AI Risk Management FrameworkNIST
Read how we source, review, update, and correct content in our editorial standards.
