The Email Click Evidence Review Framework

When this review is useful

Use this workflow when clicks jump without a corresponding business change, when a reporting filter is changed, or when two lifecycle dashboards disagree. It applies to a D2C product launch, a consumer subscription renewal message or a startup nurture send. It does not determine inbox placement, detect bots, prove a human clicked or estimate incremental revenue. If all you have is an aggregate click total, request recipient-grain evidence before attempting the partition.

Four gates before a campaign decision

1. Freeze the contract. The lifecycle owner names one send and one decision. The analyst records the delivered roster, click window, export timestamp, report filter and classification field. Finance or sales owns downstream value definitions. A multi-send unique-person report is not equivalent to a single-send report.
2. Reconcile the grain. Deduplicate stable event IDs and aggregate by anonymous recipient within that send. Quarantine out-of-roster rows, missing recipient IDs and conflicting event versions. Do not silently fix mismatches by shrinking the denominator. The analyst signs off the discrepancy log.
3. Partition before calculating. Anyone with a false flag is retained under the explicit-unflagged policy. Split retained recipients into no-true and mixed-true groups. For the remainder, all-true goes to flagged-only; any missing flag goes to unresolved. The buckets must sum to observed clicked recipients and never exceed delivered recipients.
4. Review the business evidence. The channel owner compares definition-consistent snapshots and requests consent-aware downstream evidence. An unexplained reconciliation failure means hold; an explained definition break means annotate, not declare campaign deterioration; clean evidence permits proposing a test, not automatic spend or audience changes.

Worked synthetic decision

A send reaches 10,000 recipients. The four buckets contain 240, 60, 500 and 100 recipients. Raw observed engagement is 900/10,000=9%. Explicit-unflagged engagement is (240+60)/10,000=3%. Including unresolved recipients gives 4%; the policy gap is one percentage point. Removing the 60 mixed recipients would discard valid evidence under the chosen policy.

The marketer initially proposes rewriting the offer because yesterday's report showed 9% and today's shows 3%. The analyst discovers that these are the same send under different filter definitions. The decision is to hold the creative rewrite, preserve both report snapshots and reconcile order counts separately. No business-performance change has yet been demonstrated. The next review occurs after the analyst records export coverage and the lifecycle owner confirms the filter history; no arbitrary numerical threshold is required.

Use the worksheet

Copy the worksheet above into your operating document. Attach aggregate counts and references to access-controlled evidence rather than customer email addresses. Record who can approve a reporting restatement separately from who can change a campaign. A discrepancy is resolved only when the owner can explain its origin, scope and impact, not when the two dashboards happen to show the same number.

Failure modes and limits

Missing flags are not false. Bot and unflagged recipient sets overlap. Historical data can use different classification coverage. A recipient may click on several devices or receive forwarded links, so an email identifier is not perfect person identity. Classifier errors remain possible even with a reconciled ledger. Do not turn this diagnostic into automatic suppression, customer scoring or a claim of causal lift. Review local privacy and retention requirements with the responsible team.

Primary sources checked October 11, 2026: Mailchimp bot filtering and Klaviyo bot-click documentation. Vendor settings and reporting surfaces differ. The four-bucket worksheet is original Growthcraft synthesis, not either vendor's product specification.