Confirm the denominator and source

Record the provider, complaint-feedback method, metric definition, numerator, denominator, time zone, and comparison baseline. Provider dashboards can calculate rates differently; Yahoo notes that its spam-rate denominator relates to inbox delivery, while other systems may use delivered or reported volume. Do not combine incompatible rates.

Check for delayed data, dashboard outage, duplicated events, or a sudden denominator drop. Even when measurement explains part of the spike, treat individual complaints as suppression events according to policy.

Segment before guessing

Break the affected period down by From domain, DKIM domain, sending IP/pool, provider account, campaign, template, subject, audience cohort, acquisition source, list age, frequency, geography, and link domain. Compare with a normal period.

Ask what changed: new import, partner source, automatic opt-in, frequency increase, brand name, template, re-engagement cohort, suppression synchronization, or account compromise. Authentication pass does not mean recipients wanted the message.

Contain the likely source

Pause the narrowest cohort supported by strong evidence, with authority to widen the pause if complaints continue. Stop scheduled retries and provider-side automations too. Verify that unsubscribe and complaint events propagate centrally. Revoke compromised credentials if abuse is suspected.

Do not switch domains or IPs to escape reputation. Do not delete complaint history. Those actions spread harm and destroy diagnostic continuity.

Repair recipient expectation

Remove or quarantine sources without clear permission. Correct misleading signup copy, prechecked boxes, frequency mismatch, sender identity, and unsubscribe failures. If one-click unsubscribe is required for the stream, verify the headers and POST endpoint on a real received message. Reconcile every suppression store.

Restart with gates

Choose an engaged, clearly permitted cohort and a representative message. Define acceptable complaint, bounce, and provider-response thresholds and a stop condition. Send in a controlled window and monitor by provider.

Finish with a timeline, root cause, affected scope, suppression action, control added, owner, and follow-up date. The goal is restored recipient expectation and reliable removal—not merely a dashboard rate below a published ceiling.