← All articles
Deliverability5 min read

What a Weekly Domain-Health Review Should Contain

MR
Brian Brackeen
Mailrun notes · Nov 2026

A weekly domain-health review should end with decisions somebody owns. A chart of bounces or a reputation label is useful evidence, but it does not tell the team which campaign to inspect, which configuration to repair, or what needs to be checked again.

For outbound email operations, I would keep the review small enough to repeat: a consistent reporting period, evidence separated by receiving provider, a record of changes, and a short action log. Urgent delivery failures or complaints still need attention when they occur. The weekly meeting is a place to connect the evidence and follow through on open work.

Define What the Numbers Describe

Begin each review with the sending domains, sending systems and recipient-provider groups included in the report. Record the time zone and period for each data source. If one export uses local dates and another uses UTC, reconcile the boundaries before comparing them.

Keep counts beside rates. A change in the mix of recipients or the volume sent to one provider can change an overall percentage without explaining what happened to a particular sending domain. Write down the denominator and the event being counted: attempted recipients, accepted recipients, temporary failures, final failures or recipient complaints.

Avoid treating these as interchangeable measures. In SMTP's reply-code model, a 4xx response signals a temporary failure and a 5xx response a permanent failure for the command. Preserve the full diagnostic and the command or delivery stage involved. A receiving server's acceptance of a message is not proof that it appeared in the recipient's inbox or was read.

Read Each Signal With Its Limits

A useful review sheet might contain the following entries. These are suggested review categories, not a universal health score:

EvidenceQuestion for the reviewCaution
Authentication resultsDid the intended sending path authenticate and align?A passing result does not establish recipient interest or inbox placement.
Deferrals and rejectionsWhich provider, sending path and campaign share the failure?Preserve the diagnostic before assigning a cause.
Recipient complaintsWhich messages or audiences need investigation?Provider reports can cover different populations.
Unsubscribe handlingDid suppression reach work still waiting to send?A recorded click alone does not prove the next message was stopped.
Operational changesWhat changed before the signal moved?Timing suggests a hypothesis, not proof of causation.

For Gmail, Postmaster Tools documentation describes reporting for personal Gmail accounts. Its data is not real time, uses UTC, and may be incomplete on low-volume days. Record missing or delayed observations as such, rather than turning them into a healthy zero.

Google also explains that a low user-reported spam rate can coexist with messages being filtered into spam: people have fewer opportunities to complain about messages in their inbox. That makes the report useful for investigation, not an all-clear for every recipient or provider.

Use the receiving provider's own current documentation when interpreting its signals. Do not carry a Gmail metric's meaning over to a Microsoft or Yahoo report just because the labels look similar.

Check the Sending Path Before Changing It

When a signal changes, identify the actual domain and sending system involved. Keep the visible From domain, authentication domains and sending IP or service distinguishable in your notes. This helps the person investigating follow the right path instead of changing an unrelated DNS record.

Google's sender guidelines distinguish requirements for all senders from additional requirements for bulk senders. They include authentication and, for bulk senders, DMARC and alignment requirements. Check the rules that apply to your traffic rather than using a successful send from a different system as evidence that the campaign path is configured correctly.

Make configuration changes through the person responsible for that service. A review meeting is not a reason to replace authentication records speculatively. Preserve a representative received message's headers and the relevant failure diagnostics so the administrator can reproduce the issue. Share only the details needed for the investigation, through an appropriate internal channel.

Turn an Observation Into an Owned Action

Consider a hypothetical review in which temporary failures increased at one receiving provider after a campaign change. Another provider's results look steady, and the Gmail dashboard has not yet caught up with the reporting period.

That evidence does not establish a single cause. The action could be to have the sending operator inspect the affected diagnostics and queue, while the campaign owner reviews the audience and recent message changes. Record whether affected sending should be held while the issue is investigated, who can make that decision, and what evidence is needed before resuming.

Do not respond by moving the same questionable traffic to fresh domains to work around receiver protections. If complaints or rejections point to unwanted messages, address the audience, relevance and sending practice. Changing infrastructure does not resolve that underlying problem.

For every action, capture four things: the observation, the owner, the next check, and the condition that would close the issue. An entry such as “investigate reputation” leaves too much open. An entry asking the sending operator to explain a specific set of diagnostics for a named campaign and period is something the next reviewer can follow.

Carry the Decisions Into Next Week

Start the next review with last week's actions. Confirm what changed, which evidence arrived afterward, and whether the original question was resolved. Keep unresolved work visible when a dashboard has no new data.

The useful output is a short record that another operator can act on: what is known, what remains uncertain, and who is handling it. That is more valuable than a single color attached to a domain, and it makes the review worth repeating.

Talk through your Mailrun setup.

Ask about Agent Identities for your agent or ours, or Mailrun Core for your sequencer. Our team can help you choose.

Book a call