Separate authentication from reported alignment
SPF and DKIM Alignment in DMARC Reports
Two raw authentication passes can sit beside a DMARC failure without making a report contradictory. The methods may have authenticated domains unrelated to the visible From domain. Conversely, a single reported aligned pass is sufficient for a derived DMARC pass even when the other method fails. Learning which XML fields answer which question prevents unnecessary configuration changes and misleading success rates. This guide follows the evidence from identities, through individual rows, to count-weighted summaries without pretending to perform a fresh authentication check.
Use the DMARC Report Analyzer with Files and Report files, or Paste XML and XML report. Load example fills only; choose Analyze reports explicitly. Begin with the per-report context, inspect a row, and compare its reported alignment with its raw authentication declarations. Clear ends the inspection and removes the current inputs and results.
Give each domain its proper job
Start with header_from under identifiers: it declares the author domain associated with the visible From field for that record. Then read the SPF domain and any scope supplied under auth_results, followed by every DKIM domain and selector there. These are separate identities, not interchangeable labels for one sender. SPF reports authorization of a connecting host for an SMTP identity; DKIM reports signature verification for a signing domain. Neither raw result alone states that the authenticated domain has the required relationship with header_from. Record the roles before comparing spellings, and leave missing declarations missing rather than substituting the policy domain for every blank.
For an invented billing stream, header_from is ledger.example, the reported SPF domain is bounce.route.example, and the DKIM signing domain is route.example. Both raw results say pass. That tells you what the receiver declared about those two method checks, not that ledger.example was authenticated through alignment. In row detail, inspect policy_evaluated separately. If its SPF and DKIM fields both say fail, the derived DMARC state is fail. A useful investigation note identifies the two authenticated domains and the author domain instead of saying that authentication is broken everywhere. The narrow problem may be how an otherwise expected application presents its domain identities.
The policy_published domain has another job: it identifies the domain of the reported policy. It need not be a replacement for header_from or for the SPF identity. Similarly, an envelope_from declaration is context, not permission to manufacture a missing SPF result. An empty SMTP reverse path can involve HELO-related SPF context; retain any scope actually reported rather than assuming MAIL FROM was always checked. If individual-message context is needed, the Email Header Analyzer can inspect supplied raw headers separately. It does not retrieve the original message, prove trust in a reporting service, or automatically match a message to this aggregate record.
Read policy_evaluated before deriving the DMARC state
In this analyzer, the SPF and DKIM fields under policy_evaluated are the reported DMARC-aligned authentication results. The result fields under auth_results are raw method declarations. Do not combine the two locations into a single bag of pass tokens. The derived rule is explicit: either aligned method passing makes the row pass; both aligned methods failing makes it fail; every remaining unresolved combination is unknown. An aligned pass plus a missing value is still pass. An aligned fail plus a missing value is unknown, not fail. A missing or unsupported value produces an unknown method state and a warning instead of being silently rewritten.
Suppose a row for 192.0.2.78 has count 240, aligned SPF fail, and aligned DKIM pass. All 240 reported instances contribute to the derived pass count, not 120 and not two separate groups of 240. Reverse the method values and the same rule holds. If both aligned fields are fail, those 240 contribute to fail. If SPF is fail and DKIM is absent, they contribute to unknown. This is a summary of receiver declarations, not 240 local signature tests. The source-IP summary should lead you back to these row combinations when one address contains multiple streams with different results.
Now imagine an unusual row whose raw DKIM entry says pass, while reported aligned DKIM says fail and aligned SPF is unknown. The local reader leaves the derived state unknown. It does not promote the raw pass into an aligned pass, and it does not recompute alignment from domain strings. Preserve the selected raw XML and ask whether the reported properties are incomplete or whether the receiver used context absent from the report. Successful structural parsing does not certify that every declaration is mutually consistent. If the input itself is malformed or exceeds a limit, the whole batch is rejected; that validation error is distinct from an accepted row with unknown alignment.
Use alignment modes to frame a configuration question
Strict alignment requires identical domains; relaxed alignment uses the same organizational domain. The reported aspf and adkim settings refer to SPF and DKIM respectively. These are domain relationships, not a comparison of display names, mailbox local parts, or selector names. Do not assume that taking the final two labels always finds an organizational domain. The current RFC 9989 framework includes discovery rules that this local report reader does not execute. For practical inspection, use the report to identify which relationship needs checking by an authorized administrator. It is not a reason to have the analyzer guess an organizational boundary from a convenient-looking suffix.
Consider an invented setup where mail.ledger.example and ledger.example are under the same organizational domain. An SPF pass for mail.ledger.example can support relaxed alignment with author domain ledger.example, but those domains are not identical for strict alignment. A DKIM signature with d=ledger.example has an exact-domain relationship instead. If an administrator deliberately changed aspf to strict, compare the observed outcomes for that particular stream across complete windows. Do not prescribe relaxed mode automatically: the intended administrative boundary matters. A report identifies the practical effect to investigate, while domain control and policy intent must be established through records outside this tool.
Read policy options as observed historical data. A report carrying p=none or adkim=s does not establish what DNS publishes today, and a missing field is not a fresh lookup of its default. Preserve legacy pct alongside newer np, testing, or discovery_method values when supplied, without treating them as interchangeable controls. RFC 7489 explains older DMARC reports; current DMARC is specified by RFC 9989, with aggregate reporting in RFC 9990. Older examples in provider documentation can still illustrate a legacy file, but do not turn their sampling instructions into a universal current rollout recipe. Keep the report era, observed settings, and intended change date in your notes.
Read the two result layers
Inspect a reported alignment row
Compare policy_evaluated with auth_results and retain the domain roles attached to each declaration.
Handle multiple signatures and partial repairs carefully
One row can list several DKIM results without representing several messages per instance. Suppose 600 instances have a raw pass for relay.example, a raw fail for ledger.example, aligned DKIM fail, and aligned SPF fail. The favorable relay signature does not rescue the reported alignment result. Ask the application owner whether the intended ledger.example signature was present and what the receiving system declared about it. A selector is a useful configuration reference, not an identity certificate or a message counter. The analyzer does not fetch its public key, test its signature, or diagnose a body-hash failure from aggregate declarations alone.
After an authorized signing change, a later row might retain the relay.example pass and show ledger.example pass, with reported aligned DKIM pass. That row derives pass even if aligned SPF still fails. Describe the narrower improvement: the receiver now reports an aligned passing path for this stream. Do not describe the envelope setup as repaired unless its own evidence changed. Yahoo adds a receiver-specific caution: its FAQ does not guarantee DMARC pass when multiple aligned signatures have mixed outcomes. Preserve the receiver-reported aligned field rather than using one raw signature to predict how Yahoo must decide. The general derived rule and receiver implementation evidence answer different questions.
A practical handoff names the report, window, From domain, relevant authentication domains, source IP, and weighted count, then asks one focused question. For example: why does the known reminder route at 198.51.100.94 still report both aligned methods failing after the newsletter route changed? That separates two application owners instead of broadening authorization for every source. If a forwarded or modified message is suspected, obtain authorized message-level evidence before assigning that cause. Unknown source ownership remains unknown; the tool performs no IP reputation, geolocation, or provider autodetection. An unfamiliar IP is neither proof of abuse nor sufficient grounds to add it to an SPF record.
Count message instances rather than rows or signatures
Use record.count as the weight for every row. Imagine 900 instances with aligned pass, 60 with both aligned methods failing, and 40 unknown. The total is 1000 reported instances: 90 percent pass, 6 percent fail, and 4 percent unknown when the denominator is all reported instances. If you separately calculate a known-only pass share, it is 900 divided by 960, or 93.75 percent, and must be labelled known-only. Do not let that second denominator hide the 40 unresolved instances. Averaging the three row percentages or counting their raw signatures would answer a different question and distort the operational picture.
Per-report comparisons also need weights. One window with 90 passes out of 100 and another with 1 pass out of 10 do not combine to 50 percent. Together they contain 91 passes out of 110, about 82.73 percent, provided combining their populations is appropriate. A report with zero total count has an unavailable percentage, not a measured zero-percent success rate. Zero-count rows can retain useful declared structure without contributing volume. Before comparing two windows, read their lengths, policy domains, and stream mix in the per-report view. A changing denominator can move a batch rate even when no individual sending configuration changed.
Batch identity checks prevent one narrow kind of double counting, not every repeated observation. Equal meaningful content with the same normalized reporter identity, report ID, policy domain, and begin and end is skipped as a duplicate. Different content under that identity aborts the batch. Separate reports with overlapping or touching windows warn and retain their counts; the analyzer cannot deduplicate individual messages across them. Consequently, totals are reported message instances, not unique emails, unique recipients, or the amount your organization sent. An IP recurring in multiple records is not a duplicate-message key. Inspect warnings before using any combined rate to justify a policy decision.
Keep handling, validation, and privacy separate from alignment
Disposition is not the derived DMARC state. Both aligned methods can fail while disposition none reflects monitoring or a reported local override; that is not automatically an error or malicious mail. Current RFC 9990 includes disposition pass, distinct from none, for passing DMARC with an enforcing policy. Preserve none, pass, quarantine, reject, and other values separately. None is not an inbox receipt, reject is not a complete delivery trace, and pass does not certify safe content. The summary statistics, per-report entries, source-IP grouping, row detail, and selected raw XML viewer support an investigation; they do not establish delivery, spam placement, fraud, or a human identity.
The input boundaries apply before useful totals exist: up to 10 XML, GZ, or ZIP files, each at most 4 MiB, with 16 MiB combined source and expanded-batch limits. Expanded XML must be UTF-8, optionally with a BOM, and at most 4 MiB each. ZIP allows one XML plus directories, not extra files, nested archives, or encryption. Additional caps are 5000 rows per report, 10000 per batch, depth 24, 150000 XML nodes, and 4096 UTF-16 units per leaf text value. An invalid member rejects the whole batch. Legacy no-namespace feedback, the named legacy http://dmarc.org/dmarc-xml/0.1 namespace, and the RFC 9990 namespace urn:ietf:params:xml:ns:dmarc-2.0 are supported, but acceptance is not full XSD certification.
Analysis runs locally in a 12-second worker inside an opaque-origin sandboxed standalone PHP document served no-store with connect-src none. No analysis API, DNS query, SPF recheck, or signature test is performed. Clear, hidden-page transitions, and navigation lifecycle events stop work and wipe source Files, pasted XML, and every derived result, including raw XML. There is no Copy, PDF, export, project storage, or history. These controls do not guarantee secure memory erasure, extension isolation, or protection against a compromised device or hostile parent replacing the interface. Before leaving the view, finish your bounded finding through your established evidence process: what was reported, how many instances it covers, and what remains unverified.
Questions about SPF and DKIM Alignment
Must SPF and DKIM both pass alignment for DMARC to pass?
No. One reported aligned pass suffices for the derived DMARC pass. Both reported aligned methods must fail to derive fail. If neither passes and at least one is missing or unsupported, the result remains unknown with a warning. Raw method passes cannot replace missing reported alignment.
Why does SPF pass in auth_results but fail in policy_evaluated?
The first is a raw check for its reported SMTP domain; the second is the reported aligned SPF outcome. A domain can pass SPF without aligning with the author domain. Inspect the identities and observed mode, then ask the sending administrator about the intended setup. This tool does not recompute alignment.
Should I average report percentages to obtain the batch result?
No. Sum the relevant record.count values and divide by the clearly stated message-instance denominator. Keep unknown counts visible and review duplicate or overlap warnings. A zero denominator makes the rate unavailable. Signature counts, row counts, and unweighted report averages are not substitutes for message weights.
Continue with another DMARC-report guide
Read an aggregate report in the DMARC Report Analyzer & XML Viewer or choose a related task below. Explore more practical utilities in all free tools.
Use the right denominator
Review aligned message counts
Inspect reported pass, fail, and unknown instances while keeping disposition and unresolved questions separate.