Key takeaways
- A conversion occurrence is not the same as every event reporting it.
- Attribution and fraud risk are independent dimensions.
- Value should come from governed business data, not a risk multiplier.
- Late outcomes require recalculation with history.
Canonicalize before evaluating
A browser tag, server event, advertising import, payment event, and CRM update may refer to one real-world conversion. Establish the occurrence and deduplication rules, then retain each source event and its reconciliation state.
This avoids counting duplicated events as several outcomes or deleting evidence needed to explain a later correction.
Keep the decision dimensions separate
Validity asks whether the occurrence is accepted under measurement rules. Risk describes suspicious evidence. Qualification applies business criteria. Attribution assigns acquisition credit. Value records a governed amount. Each may change on a different timeline.
- Attach visitor and session evidence with eligibility context.
- Preserve internal and provider attribution separately.
- Mark pending outcomes instead of assuming zero.
- Version qualification and risk decisions.
Reconcile downstream outcomes
Refunds, chargebacks, cancellations, CRM qualification, and offline sales can refine interpretation. Apply them as dated evidence and corrections, not as silent edits to the original conversion.
Limitations
What this guide does not claim
Conversion definitions, attribution models, refund windows, and downstream access vary by business. Risk evidence cannot independently determine financial value.
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.