One Inbox on a Domain Isn't Getting Replies: Recreate It, Don't Cancel It

The scenario

Three inboxes on one domain, the same campaign, the same list, the same copy. Two of them are booking replies at the rate you expect. The third has sent its share and heard nothing for a week. The natural request is to cancel that one and keep the other two, and it is the request we get most often from customers in this spot.

It feels like the tidy fix: remove the part that is not working. But it treats the inbox as the unit of reputation, and for cold email the unit is mostly the domain. Once you read the situation that way, the three-inbox domain is not one broken part and two good ones. It is a measurement: three senders on the same domain, and the split between them tells you where the problem lives.

Why the domain is the unit receivers judge

Every inbox on a domain signs with that domain’s DKIM key, passes or fails that domain’s SPF and DMARC, and carries the domain in its From address and its links. When Google or Microsoft decide what to do with a cold email, the domain is the identity they have the most history on, and it is shared by every mailbox on it.

You can see this in the refusals receivers send. Google’s permanent refusal for bad reputation does not name a mailbox; it says the sending domain has a very low reputation. In our own study of 3,415,449 sends, that refusal was concentrated on 127 of 3,493 domains rather than spread across mailboxes, and once a domain drew it, it drew it for every inbox on that domain.

Mailboxes are not invisible to receivers, and an individual account can be restricted, suspended or filtered on its own. But that is the exception layered on top of the domain’s standing, which is exactly why the split between mailboxes on one domain is so informative.

First, make sure the silence is real

Cold email reply rates are small numbers, and small numbers produce long runs of nothing by pure chance. Before deciding a mailbox is broken, check how likely it is that a mailbox that is perfectly fine would have got zero replies from the sends it actually made:

At 15 sends a day, a Google inbox takes about a week to reach 100 sends. So a week of silence at a 2% reply rate is roughly a one-in-eight event for an inbox with nothing wrong with it. Before acting, check the comparison is fair:

• •Similar sends. The sequencer may have given the quiet inbox fewer sends, or paused it after an error. Compare sends, not days.
• •The same campaign and list slice. Rotation hands each sender a different slice of the list. One slice heavy with security gateways or stale addresses can go quiet on its own.
• •The mailbox is still connected. A revoked token or a password reset stops sending or reading without anyone noticing. Check the sequencer for connection errors.
• •Replies are not landing somewhere else. Reply handling, forwarding and the unified inbox can hide one mailbox’s replies. Log in to the mailbox and look.

Read the pattern

With three inboxes on a domain, which is how our prewarmed Google and Microsoft inboxes are sold, the split reads like this:

On an Azure tenant, where one domain can hold 25, 50 or 100 inboxes, the same rule holds as a majority: a handful of quiet mailboxes on a tenant where most reply are mailbox problems; most of the tenant going quiet is a domain problem.

Two out of three: why one mailbox goes quiet on a healthy domain

The domain is fine, so it is something that belongs to that one account. In our experience it is usually one of these:

• •Spam reports concentrated on that sender. Each inbox works its own slice of the list. If its slice reported more, the receivers’ view of that sender worsens even while the domain stays in good standing.
• •A restriction on the account. Google suspends or rate-limits a user that sends like a spammer, which shows as a suspended user in the Workspace admin console or a “daily user sending limit exceeded” (550 5.4.5) bounce. Microsoft puts outbound-spam senders on its restricted list, visible as Restricted entities in the Defender portal, and their mail bounces with “Access denied, bad outbound sender” (550 5.1.8).
• •Filtered on its own. Mail from that address is accepted and goes to spam while its siblings land: what operators call shadow-banned. Nobody outside the receiver can see a per-mailbox score, so it is an inference from the pattern, and a placement test is how you check it.

How to check it

• •Run a placement test from the quiet inbox and from one healthy sibling with Burned Inbox, our free test. Spam on one and inbox on the other settles it.
• •Look at the account in the Google Workspace admin console or the Microsoft Defender portal for a suspension or restriction.
• •Paste that mailbox’s bounces into the bounce analyzer, and the headers of a test send into the header analyzer, to see what the receiver said.

What to do: recreate it under a different name, then warm it

Not cancel, and not rename. Cancelling gives up a slot on a domain that is working, which is the most valuable kind of sending capacity you have. Renaming keeps the same account: in Google Workspace a renamed user keeps its history, and its old address carries on as an alias, so whatever happened to the account comes with it.

A new mailbox under a different name starts without that history, on a domain receivers already trust. It still has no history of its own, so it warms before it takes campaign volume. Start it below the cap and ramp it, and do not push the two healthy inboxes above their cap to make up the difference in the meantime: that is how a two-out-of-three domain becomes one out of three.

One out of three: the domain is the problem

When most of a domain has gone quiet, the one inbox still replying is the lucky one, not proof the domain is fine. Recreating mailboxes on that domain recreates them into the same problem. Work the domain:

• •Rule out broken setup. A silently broken SPF or a rotated DKIM key looks like a burned domain from outside. Run it through the deliverability checker first; it fixes in an afternoon.
• •Check the blocklists. The blacklist checker and the SURBL checker. Our tools do not query the Spamhaus DBL, which needs a paid key, so check that one at check.spamhaus.org.
• •Read the bounces for Google’s domain-reputation refusal. A domain drawing it is not warmed back; the study has the line where retiring it is the cheaper answer.

Retire it, or move it to the other provider

Retire the domain when it is listed, refused on reputation, or has failed a recovery cycle. Replace it with a new domain on a .com, and its inboxes with it. Recover, rest or retire is the full decision.

Move it to the other provider, Google Workspace to Microsoft or the reverse, when the domain checks clean but the failures sit with one provider’s sending. The domain keeps its age and its authentication; its mailboxes are rebuilt on the other platform and warmed there. Then send it to the recipients that platform should be sending to. Google-hosted recipients are sent from Google Workspace; everyone else (Microsoft 365, other providers and gateway-protected domains) from Azure. The migration checklist covers moving without losing replies.

How this works at WarmInboxes

• •Cancellation is by domain. Cancellation in the WarmInboxes app is by domain: a domain is cancelled together with all of its inboxes, and a single inbox on a domain is not cancelled on its own.
• •One quiet inbox on a working domain. One inbox underperforming while the others on its domain reply is a problem with that mailbox, not the domain: we recreate it under a different name and warm it again rather than cancel it. Message support on WhatsApp or raise a request in the app, with the placement test or bounces that show it.
• •A failing domain. Most of a domain’s inboxes underperforming is a problem with the domain or its provider: retire the domain, or move it to the other provider (Google Workspace to Microsoft, or the reverse). Support diagnoses it with you first, because a broken record or a bad list slice is cheaper to fix than a domain.
• •In the first week. Under the 7-Day Placement Replacement Guarantee, an eligible inbox that fails our placement standard in the first seven days after delivery is replaced at no charge, on the evidence you send. After the first week, support diagnoses the problem with you and advises whether a paid replacement or another fix is the right answer.

The checklist

• •Confirm the silence: enough sends for zero replies to mean something, compared with siblings on the same campaign and list.
• •Confirm the mailbox is connected and that its replies are not landing somewhere else.
• •Count how many inboxes on the domain are replying.
• •Most replying: check the quiet mailbox (placement test, admin console, bounces), then recreate it under a different name and warm it.
• •Most quiet: check the domain (authentication, blocklists, reputation refusals), then retire it or move it to the other provider.
• •Every domain quiet at once: it is not the infrastructure. Work the list, the copy and the volume.

For a problem that does not fit one of these, describe it to the WarmInboxes Free MCP, which runs the checks in order and says when new infrastructure is, and is not, the answer.