Email Complaint Rate Checklist: Compare the Right Denominators

COMPLAINT-RATE EVIDENCE CONTRACT — original Growthcraft synthesis
Decision and anonymous sender scope:
Provider / dashboard / domain grouping:
UTC dates / extraction timestamp / freshness:
Source A metric name / numerator / denominator / exclusions:
Source B metric name / numerator / denominator / exclusions:
Count unit and duplicate-report handling:
Raw exact counts available? Provenance or explicitly unavailable:
Missing, privacy-suppressed or revised days:
Rounded display precision:
Daily results retained separately from pooled summaries:
Illustrative threshold, basis points, source and checked date:
Synthetic sensitivity example (not provider reconstruction):
Known mismatches / unsupported claims to remove:
Analyst / lifecycle owner / deliverability reviewer:
Next bounded investigation / review date / stop condition:
No sending or configuration change authorized by this worksheet.

Use this when email reports disagree

This original editorial synthesis helps lifecycle marketers and founders explain why two complaint percentages disagree before blaming a tool or changing a campaign. Use it when you can inspect definitions and snapshots from the relevant reports. It is not a sender-certification checklist, legal assessment or inbox-placement forecast. Examples are synthetic; Growthcraft Editorial prepared them with AI assistance, without claiming Akshay's personal review or client results.

A useful review starts with a decision: are we comparing the same population, or looking at complementary signals? Do not use the framework to manufacture a single reassuring percentage from incompatible sources. If only a rounded number is available, record that limitation. Do not reconstruct private provider counts by guessing an inbox rate.

1. Write the contract before copying the number

The analyst records the provider, dashboard, domain grouping, observation window, timezone, extraction time and metric definition. The unit should be explicit: messages, recipients and complaint reports are not interchangeable. A person receiving three messages can be one recipient and three possible message-level complaint events. A repeated webhook can be two reports about one message.

Retain the original source label and a reference to its approved internal export. Public worksheets need anonymous aggregate facts only. Keep raw addresses, message bodies and account tokens inside controlled systems. If a report has no exact numerator or denominator, leave those fields unavailable rather than filling them from another dashboard.

2. Compare eight dimensions

Compare population, unit, time attribution, timezone, authentication/domain scope, deduplication, exclusions and refresh status. A difference in any one can make two percentages non-equivalent. Do not assume that a provider domain grouping matches a campaign ID in your sending platform. Ask the owner to document the mapping, including streams that share a sending identity.

Keep a table with one row per source and date. Use a separate status for missing or suppressed values. A value of zero means an observation exists under its declared definition; a missing observation means you do not have that evidence. A delayed refresh should trigger a later check, not an automatic campaign-quality conclusion.

3. Separate reconciliation from sensitivity

Reconciliation explains observed source differences. Sensitivity asks what a compatible hypothetical count would imply. Label those tasks separately. The companion calculator supports the second without claiming to recover a provider's hidden denominator. Its integer threshold calculation avoids classifying a rounded display as below a strict boundary.

Use decision criteria chosen by the responsible owner, not a universal score invented by this worksheet. If the team cannot establish comparable scopes, the output is a mismatch list and an evidence request. If it can, compare exact fractions and preserve the daily observations behind any aggregate.

4. Work a synthetic discrepancy

A synthetic ledger has two complaint-marked messages among 1,000 eligible messages. The fraction is 0.20%. A separate broad delivery total is 4,000; dividing the same two by it gives 0.05%. Neither arithmetic operation is wrong, but the denominators answer different questions. The 4,000 must not be substituted into an inbox-population metric merely because it is easier to export.

At an illustrative 30-basis-point boundary, three complaints among 1,000 reach 0.30%; two are the largest whole count strictly below it. One additional complaint reaches the boundary, with the population fixed. This is not a spare-complaint allowance, forecast or permission to send. At only 250 eligible messages, one complaint is already 0.40%, illustrating discrete-count sensitivity rather than a different policy.

5. Assign a bounded next action

The analyst resolves the count and snapshot contract. The lifecycle owner reviews the relevant consent, expectation and campaign context through approved internal processes. The deliverability specialist interprets provider-specific status and any operational proposal. An unresolved count mismatch blocks a claim that the reports have been reconciled; it need not block responsible investigation using each report on its own terms.

Record the question, owner, review date and required evidence. For the synthetic example, the next action is to verify whether the broad delivery total includes messages outside the eligible population, not to increase send volume until the percentage falls. Do not rotate identities, hide feedback, remove security controls or treat this worksheet as authority to alter infrastructure.

Failure modes and limits

Watch for averaging daily percentages, rounding before comparison, treating late reports as duplicates without checking keys, joining a message count to a recipient count, and mixing date-of-send with date-of-report. A mathematically correct fraction does not validate its source. This framework cannot establish causal campaign effects or prove future inbox placement. Preserve unresolved definitions in the decision note rather than disguising them with extra decimal places.

Primary sources and boundaries

Checked October 26, 2026; publication/update dates not stated on these living pages. Google Postmaster dashboards documents its qualified inbox population and missing low-volume data. Yahoo Sender Hub FAQ distinguishes inbox-based complaint rate from broader delivery reporting. Neither source supplies your exact underlying counts or endorses this calculator.