Start with three different questions

Delivery usually answers whether the receiving mail system accepted responsibility for the message during SMTP. Placement describes an observation after acceptance, such as inbox, spam, another tab, or missing from a monitored mailbox. Engagement describes a recipient action such as a click, reply, or complaint. These events happen at different points and should not be collapsed into one “delivered” metric.

A 250 2.0.0 response is useful evidence that the receiving server accepted the message. It does not identify the final folder and it does not guarantee that a person saw the mail. RFC 5321 also allows a receiving system to generate a later non-delivery notification if delivery fails after acceptance.

Build a small evidence table

For one authorized test, record facts without filling gaps:

Stage Example evidence Safe conclusion
SMTP handoff 250 2.0.0 accepted at 10:04 UTC The named receiver accepted responsibility at that time
Mailbox observation Message seen in spam at 10:05 UTC This monitored account placed this copy in spam
Engagement No click event after one hour No tracked click was observed; it does not prove the message was unread

Keep the provider, account type, message identifier, and timestamp beside each observation. A seed mailbox result from one provider cannot represent every recipient at that provider, and an open event can be affected by image blocking or automated privacy features.

Diagnose the stage that actually failed

If SMTP rejected the message, begin with the response code, responding host, and the sending identity. If SMTP accepted it but a controlled seed shows spam, inspect authentication, reputation signals, recipient expectations, and recent changes. If placement appears healthy but engagement falls, investigate audience, frequency, and content expectations before changing DNS.

Do not use “100% delivered” when the platform only measured acceptance. Prefer language such as “98.7% accepted by recipient servers; placement was observed in 18 authorized seed accounts.” That sentence says exactly what was measured.

Verify your report

Choose three recent events and trace each from sending log to mailbox evidence. Check that every chart uses one definition consistently, time zones are explicit, and unknown or missing observations remain unknown. Your report is ready when another operator can reproduce the classification without guessing what “delivered” meant.