This DMARC readiness checklist records the evidence for a proposed policy change on one domain: what evidence you have verified, which traffic paths remain unresolved, and who owns each decision. A passing sample from a daily notification stream tells you nothing about whether an unobserved monthly batch job will survive the same policy transition.
Filling in the path rows first
Begin by compiling your approved sender catalog, business owners, mail and DNS administrators, active policy records, proposed enforcement scope, aggregate report references, and raw sample headers. Reference specific ticket numbers, log entries, or report IDs; never paste passwords, credentials, or customer message bodies into the sheet. Copy the blank worksheets and assign a dedicated row to each distinct combination of sending service, tenant account, scheduled job, and delivery route. When return paths, schedules, or intermediary relays differ, track them on separate rows; one row cannot hold two different sets of criteria. Mark fields as unknown whenever evidence is missing.
The worksheet itself is a recommended working method, not a requirement of the standard. Use these six columns to evaluate each path before making policy changes:
-
Path and cadence: identify the sending service, the account or tenant, the specific job, the Author Domain, direct versus indirect routing, and whether traffic runs continuously, monthly, or seasonally.
-
Owners and approval: name the accountable business owner and technical lead, linking to a dated business authorization. Appearing in a report does not establish approval.
-
Evidence and coverage: reference server logs, aggregate report generators and report IDs with UTC timestamps, sample IDs, and the receiving mail administrator. Detail which recipient domains and business cycles were observed and which remain unexamined. Before relying on an
Authentication-Resultsheader, verify where it came from, since a pasted header alone is insufficient (header trust, RFC 8601). Suspicious telemetry demands corroboration because aggregate report data can be forged. -
Authentication and handling: capture the visible
header.from, the SPF MAIL FROM domain with its raw evaluation, the DKIM signing identityd=and selector with their raw evaluation, alignment modes and outcomes, overall DMARC evaluation, and discovered policy. Keep raw authentication distinct from alignment and disposition, logging reported local policy overrides and missing parameters explicitly. -
Criterion or gap: define the exact telemetry required to clear this path for policy tightening, including low-frequency batches and forwarded mailing list routes. Justify why your observation timeframe fits the real-world business cadence; RFC 9989 treats these as cadence-sensitive decisions, and fitting the window can take months. An absence of aggregate reports never proves that no mail was sent, because mailbox providers can opt out of reporting entirely.
-
Decision and follow-up: specify the next action, the assigned owner, and the review trigger. Mark HOLD for incomplete evidence, REPAIR for legitimate traffic experiencing authentication or alignment failures, and READY FOR [target policy] once all criteria pass; these are worksheet labels, not standard terms. Legitimate unauthenticated or unaligned streams MUST be remediated before enforcement (required remediation, RFC 9989).
| Path and cadence | Owners and approval | Evidence and coverage | Authentication and handling | Criterion or unresolved gap | Decision and follow-up |
|---|---|---|---|---|---|
Which policy scope does the change affect?
Document the Author Domain and discovered Policy Domain, applicable p, sp, and np tags, and every inherited subdomain the policy change affects. Modern policy discovery relies on the DNS Tree Walk algorithm, prioritizing the Author Domain’s own valid record. An older aggregate report displaying discovery_method=psl documents the legacy Public Suffix List method used by that specific receiver. When the optional discovery_method tag is missing, its absence does not confirm whether the receiver used the DNS Tree Walk or a legacy PSL lookup (reported discovery method).
Sources checked: 14 September 2026; receiver adoption was not tested. The legacy pct percentage tag was removed in RFC 9989, so never assume publishing pct=25 restricts policy enforcement to a quarter of your outbound traffic. For non-passing messages, the testing flag t=y requests testing with the expectation of one policy level below the specified one: reject → quarantine, and quarantine → none. The flag alters neither none handling nor aggregate report dispatch, and default t=n requests the published policy (testing semantics). Receiver discretion remains in how the policy is handled, so none of this is a safety or inbox guarantee.
Publishing p=reject carries strict protocol prerequisites under RFC 9989: domains MUST NOT rely solely on SPF for DMARC success and MUST apply valid DKIM signatures to all outbound mail. Domains hosting human users who participate in public discussion groups SHOULD NOT publish reject policies. If an organization chooses to enforce reject on such a domain regardless, it SHOULD measure user disruption across aggregate telemetry, spending at least one month at p=none followed by equal duration at p=quarantine while monitoring disposition trends. Senders opting for reject SHOULD restrict mailing list participation and/or warn users that their participation may be hindered (reject and mailing-list conditions). Escalate this branch to the domain owner; its timing is not universal. The guide explains the mechanics behind the forwarding and mailing-list problem.
Signing off the whole scope
Complete this scope summary card only after finishing every individual path row. Marking scope as READY authorizes preparation under internal change management; it is not a direct instruction to edit production DNS. Preserve the previous approved values before approval, so that recovery has something to return to. For recommended rollout staging, review how to roll out DMARC without breaking your email.
| Scope-level decision field | Fill in before approval |
|---|---|
| Author/Policy Domains, inherited scope; current/proposed values; user/list use | |
| Observation window, cadence justification and receiver/reporting gaps | |
| Success evidence, stop conditions and escalation contact | |
| Previous approved policy/reference; recovery owner, trigger and closure evidence | |
| Decision, unresolved rows, approver, date and recheck trigger |
A synthetic example: a passing sample, a domain on hold
This walkthrough uses synthetic test data: no messages were sent, no DNS was queried and no reports were tested. The evidence summary supplies every input; it is not XML or full message source.
Both Author Domain and discovered Policy Domain resolve to ship.example.com. The active baseline record DNS-41 publishes v=DMARC1; p=none; adkim=s; aspf=s; t=n; rua=mailto:dmarc@ship.example.com. The proposed rollout updates only p, shifting from none to quarantine. The stipulated scope review finds no additional inherited sending domain. This transactional domain has no individual user mailboxes and never sends to mailing lists, though third-party forwarding has not been measured.
Application catalog INV-4 lists Dispatch account D4, which sends daily alerts and a monthly reconciliation batch on the last business day of each month. Authorization ticket APP-12, signed on 31 August 2026, approves both streams. Aggregate reports R-01 through R-07 from synthetic receiver receiver.example.net span 1 September 2026 00:00 to 8 September 2026 00:00 UTC. They report 168 daily notices matching LOG-D4, with zero monthly reconciliation events logged. This telemetry validates only this single receiver path; visibility across other destinations remains unconfirmed.
Direct-route sample H-07, received 7 September 2026 at 12:00 UTC, carries authentication results assumed verified as trusted by the recipient administrator. The header records header.from=ship.example.com, an SPF identity using MAIL FROM bounce.example.org with raw pass, and a DKIM identity using d=ship.example.com and selector dispatch1 with raw pass. Under strict alignment, SPF fails because the return path does not match the From domain, while DKIM passes on the matching signature domain. Because a single aligned authenticated identity satisfies the message pass rule, DMARC evaluates as pass.
Aggregate report tallies corroborate these findings, logging policy_evaluated/spf=fail and policy_evaluated/dkim=pass. Receiver disposition and local policy overrides are absent from this report summary, and the optional discovery_method attribute was omitted, leaving receiver Tree Walk adoption unknown. Either way, aggregate dispositions cannot show inbox placement.
| Path and cadence | Owners and approval | Evidence and coverage | Authentication and handling | Criterion or unresolved gap | Decision and follow-up |
|---|---|---|---|---|---|
D4 daily notices; ship.example.com; direct | Dispatch lead; Messaging engineer; APP-12, 31 August 2026 | R-01–R-07, LOG-D4: 168 notices; trusted H-07; period above | Strict modes; SPF raw pass/alignment fail; DKIM raw pass/alignment pass; DMARC pass; current none; handling unknown | Daily authentication criterion met only; monthly job and recipient/indirect coverage unresolved | Domain HOLD; Dispatch lead supplies monthly-job evidence; Messaging engineer resolves recipient routes |
The local criterion requires attributable, trusted aligned evidence for both authorized jobs, a justified observation window with recipient and indirect coverage, and resolved legitimate failures. Open an independent HOLD row for the monthly reconciliation stream, as zero observations cannot satisfy release criteria. On 8 September 2026 the Domain owner logs HOLD against the proposed quarantine change. The review happens when the missing evidence arrives; the owner check-in on 1 October 2026 is a chosen date, not a standard waiting period. Recovery owner: Messaging engineer; rollback baseline: DNS-41.
When does the decision stop, and when does it reopen?
Assign HOLD whenever ownership records, domain scope, or empirical evidence show gaps. Direct each unresolved item to the appropriate specialist: business owners for service authorization, DNS administrators for subdomain scope, and mail engineers for header verification and reporting gaps. Assign REPAIR to legitimate streams encountering authentication or alignment failures, requiring fresh telemetry before re-evaluating enforcement.
In the example, two things would trigger a stop after a future change: a confirmed legitimate D4 authentication regression, or an inability to assess the agreed critical recipient paths. A halt freezes policy progression, activates the approved rollback procedure with the Messaging engineer, and preserves diagnostic headers before and after changes. Apply the same pattern to your own recorded triggers and recovery owner. Recovery stays pending until the agreed received evidence and subsequent report comparisons support closure; reports can contain old or mixed dispositions during policy transitions, so a successful DNS edit does not establish recovery.
Optional reading: the sender inventory procedure and the DMARC enforcement procedure. The worksheet above can be completed without either.
Frequently asked questions
Does appearing in a DMARC report mean a sender is approved?
No. Appearing in a report does not establish approval. Name the business and technical owners and attach a dated authorization for that use.
Do missing DMARC reports prove that no mail was sent?
No. Missing reports do not prove traffic was absent because receivers can opt out of generating telemetry. Document which recipient paths and sending cycles were verified and which remain unobserved.
Does pct=25 still limit enforcement to a quarter of traffic?
RFC 9989 removed pct, so do not assume that pct=25 still limits enforcement to a quarter of traffic. Sources were checked on 14 September 2026, and receiver adoption was not tested.
What does the DMARC t=y flag request?
For failing messages, t=y requests testing with the expectation of one policy level below the specified one: reject → quarantine, and quarantine → none. It changes neither none nor report generation, and the default t=n requests the specified policy.
Does READY mean you can edit the DNS record?
No. READY permits preparation under local change control; it does not instruct a DNS edit. Preserve the previous approved values before approval, so that recovery has something to return to.


