A TLS-RPT report shows which of the reported SMTP connections failed their transport checks, so that you can investigate the policy and the endpoint it names. The total-successful-session-count, as defined in the RFC 8460 session counter specification, tallies TLS handshakes that met your policy. It doesn’t tell you whether an email reached the inbox, whether a message was accepted, which customer sent it, or whether content stayed encrypted end to end.
Read the reporting organization, the report-id, the date-range, and each block inside policies together, because session counts only make sense alongside who sent the report, what time window it covers, and which exact policy string the sender evaluated (report fields, RFC 8460 section 4.4). If a report lands in your inbox by email, confirm that it carries a valid DKIM signature from the reporting domain before trusting the numbers. Under RFC 8460 section 3, receivers must ignore reports that fail DKIM validation. That signature verifies the report itself, not the messages described inside its counters.
A synthetic TLS-RPT report, annotated
Here is a sample report to illustrate the structure. The hostnames, IP addresses, and counts are examples rather than real production data. It uses the mx-host array format specified in the RFC 8460 report schema.
{
"organization-name": "Synthetic Sender",
"date-range": {
"start-datetime": "2026-09-10T00:00:00Z",
"end-datetime": "2026-09-10T23:59:59Z"
},
"contact-info": "tls-report@example.net",
"report-id": "synthetic-20260910-01",
"policies": [
{
"policy": {
"policy-type": "sts",
"policy-string": [
"version: STSv1",
"mode: testing",
"mx: mx1.example.com",
"max_age: 86400"
],
"policy-domain": "example.com",
"mx-host": [
"mx1.example.com"
]
},
"summary": {
"total-successful-session-count": 980,
"total-failure-session-count": 20
},
"failure-details": [
{
"result-type": "certificate-expired",
"sending-mta-ip": "192.0.2.40",
"receiving-mx-hostname": "mx1.example.com",
"receiving-ip": "198.51.100.25",
"failed-session-count": 14
},
{
"result-type": "certificate-not-trusted",
"sending-mta-ip": "192.0.2.40",
"receiving-mx-hostname": "mx1.example.com",
"receiving-ip": "198.51.100.25",
"failed-session-count": 10
}
]
}
]
}
When matching a report to server changes you made, look at the timestamps inside the JSON rather than when the email arrived. This sample spans midnight to 23:59:59Z on 10 September. RFC 8460 recommends 24-hour UTC aggregation and permits delivery delays, so don’t assume that every report has the same boundaries or that it shows up right after midnight.
The policy-string field tells you what policy the sending server evaluated during the handshake, which might not match the file on your web server right now. Senders may keep using a cached policy until it expires (cached-policy selection, RFC 8461 section 5.1). Keep that string in mind during your investigation, and check the MTA-STS guide chapter if you need a refresher on policy deployment. In this sample payload, mode: testing allowed messages through despite failed TLS handshakes. Once you switch to enforce, RFC 8461 section 5 requires conforming senders not to deliver to hosts that fail validation. In either mode, these counters reflect TLS connections, not individual inbox deliveries.
| Evidence in the sample | What it means and what to do next |
|---|---|
980 successful and 20 failed sessions | The reported failure share is 20 / (980 + 20) = 2%. That is a connection-level ratio for this policy and this period, not a percentage of failed messages. |
certificate-expired: 14 | Check the certificate expiry on the named receiving endpoint for the report’s period. |
certificate-not-trusted: 10 | Look at the certificate chain and at how the sender assessed trust; the label alone doesn’t point to one exact cause. |
| Both rows name the same host and IP | Start the investigation at that endpoint. The JSON gives no per-session IDs that would let you match the two rows. |
The two certificate labels have different diagnostic scopes. Expiry is a specific condition, while certificate-not-trusted can cover several certificate validation problems (certificate failure types, RFC 8460 section 4.3.1).
In this example scenario, assume ten sessions failed on expiration alone, six on trust alone, and four on both. That accounts for 10 + 6 + 4 = 20 distinct failed sessions, even though the categories sum to 14 + 10 = 24. The JSON itself never reveals how individual sessions overlap; that breakdown is only an illustration of why detail rows exceed the total. Because RFC 8460 section 4 allows overlapping failure categories and multiple policy objects, never sum the detail rows or policy totals and assume you are counting unique rejected messages.
What to investigate, and when to hold the mode change
Start by comparing the MX named in the report with the policy that was applied, then run a live test on that endpoint. MTA-STS evaluates policy matching and server TLS identity separately. On the certificate itself, inspect the validity dates, the chain back to a trusted root, and the Subject Alternative Name matching the MX hostname (certificate validation, RFC 8461 section 4.2).
Save the report, note the root cause you suspect, check endpoint logs from that same time window, and verify the server repair before touching your policy mode. Then wait for a fresh report covering traffic after the fix to confirm clean handshakes. Hold the mode change if failures remain unexplained, if endpoint evidence is missing, or if the next observation can’t be matched to the intended policy and to the observation window after the repair. A later clean report on its own cannot reconstruct the earlier incident. Basis for the repair check, RFC 8461 section 4.2
Does every provider send TLS-RPT reports?
Microsoft notes (as of 15 September 2026) that Exchange Online sends outbound TLS-RPT reports (Exchange Online documentation). That is a statement about one provider; it does not establish that every provider participates, or that all of your traffic is covered. Even providers that do participate can omit transient network disconnects and TCP timeouts.
An absent report means you have an observation blind spot, not flawless TLS transport. Before treating quiet days as an improvement, check which providers sent reports and which dates they covered. When policy-type reads no-policy-found, the sending server could discover neither a DANE nor an MTA-STS policy for your domain; that label alone doesn’t mean the TLS handshake failed (meaning of no-policy-found, RFC 8460 section 4).
If you need background on setting up reporting records, start with the TLS-RPT guide chapter. Whenever you review incoming reports, write down three things: the mail server problem you suspect, the endpoint log evidence you still need, and the check you’ll run before changing your transport security policy.
Frequently asked questions
Does a TLS-RPT report show whether my messages were delivered?
No. The session counters track TLS connection attempts that satisfied or violated transport policies; they do not indicate whether an email was accepted, placed in the inbox, or what happened to individual messages.
Why do the failure-details counts add up to more than the total failed sessions?
RFC 8460 allows overlapping failure categories, meaning a single connection can be flagged for multiple conditions such as an expired certificate and an untrusted issuer chain.
What does policy-type no-policy-found mean?
The reporting server indicates that it could discover neither an MTA-STS nor a DANE policy for your domain. By itself, this status does not mean TLS transport failed.
Does a missing report mean there were no failures?
No. A missing report points to a monitoring blind spot rather than clean delivery. One provider’s documentation does not establish that every provider participates, and even a system that does report need not include transient network errors or TCP timeouts.
When should I hold the MTA-STS mode change?
Hold it while the failure is still unexplained, while endpoint evidence is missing, or while the next observation can’t be matched to the intended policy and to the observation window after the repair. A later clean report on its own cannot reconstruct the earlier incident.


