Since 2024 Google and Yahoo have published rules for anyone sending in volume, and the penalty for breaking them lands on the whole domain, not on the one campaign. A warning means the domain is close to the line. This page says what to do in the first week, and then what SealDeal does so it does not happen from inside the product.
Not slow it down. Stop it. Complaints lag their sends by hours or days, so the rate you see today reflects mail that already left; anything still scheduled adds to it.
Google's requirements and Yahoo's are linked above. Check SPF, DKIM and DMARC on the exact domain that sends, and that unsubscribes are one click and honoured within two days. Most warnings have one of those underneath.
A complaint is a person telling you they never wanted the mail. Remove the list segment they came from, not just the address. If the list came from a tool that sent on its own, assume the whole list is suspect until a person has looked at it.
When you resume, resume at tens a day, to people you can name a reason for writing to, from a domain whose authentication you have just checked. Grow only while the complaint rate stays flat. That is what the ramp below does, and you can do it by hand with any tool.
Eight brakes, in the order a send meets them. Each names the part of the product that applies it, so you can ask about the mechanism rather than the adjective.
Gmail, Microsoft 365 or your own SMTP relay, per rep. There is no shared sending pool, so the reputation you build is attached to your domain and nobody else's bad week can spend it. A managed transport exists for teams without a mailbox to connect, and it carries the lowest cap of all.
email-provider.ts, send modes
Every draft waits in a queue with its sources beside it. Review is the default and stays on until an admin turns it off for one named rep and one type of action. The cheapest deliverability control is a human reading the mail, and the product enforces it rather than recommending it.
approval queue
Unsubscribes, complaints and bounces go on a suppression list at the address or the whole-domain level, and the check runs at the moment of sending, not at list upload. A cross-customer network of hashed addresses adds a second layer: an address that has bounced for five or more other SealDeal customers is blocked before your first attempt, one that has bounced for two is flagged, and nobody learns who else contacted it because only a hash is stored.
suppression-check.ts, platform-suppression.ts
Connecting a mailbox enrols its domain in a warmup ramp automatically. The ramp is a volume schedule, not a network of seeded inboxes trading fake mail, and it watches itself: if the last week's bounce rate passes 5% or complaints pass 0.3% it steps the cap down instead of growing it.
warmup.ts, warmup-enroll.ts
200 a day per rep on the managed transport, 2,000 on your own SMTP relay, and the provider's own limits on Gmail and Microsoft 365. On top of those an aggregate cap per sending domain, so ten reps under their own limits cannot add up to a blast. When a send is held the queue names which brake held it, warmup or cap or complaints, because "held" on its own is not something you can act on.
send-budget.ts, reputation-budget.ts
Over a 24-hour window, once there are at least 500 sends and at least 2 complaints, a complaint rate at or above 0.1% holds every queued send for the organisation. The floors exist because one stray complaint on a small day is noise, not a signal. An admin sees the hold on the Operations page and releases it deliberately.
complaint-throttle.ts, observability
The same hold, on bounce rate, at 2% over the rolling window. A recipient domain that does not exist is classed as a hard bounce after the retry window closes rather than retried forever, so a bad list is found on the first day and not the tenth.
deliverability-health.ts, non-delivery-report.ts
Campaign mail carries the List-Unsubscribe header the bulk-sender rules require, plus a signed unsubscribe link and the physical address in the footer. Sending is refused if the address is not configured, rather than sent without it.
email-provider.ts
What none of this does is tell you whether the mail is any good. A domain can be perfectly healthy and the emails still ignored. That is what the holdout is for.
Google asks bulk senders to keep the spam rate reported in Postmaster Tools below 0.10% and never to reach 0.30%. Yahoo publishes its own requirements and its own dates. Both require SPF, DKIM and DMARC alignment and one-click unsubscribe. Each provider's page is linked above; read the one you send through rather than a summary of it, including this one.
No software can. What it can do is stop the burning: the holds and caps above mean a second bad week is not possible from inside the product. Recovery is a matter of low, clean volume over weeks, from a domain whose authentication is correct, to a list that has been cleaned. The ramp is built for exactly that, and the holdout will tell you whether the recovered domain converts.
Many teams do, so that a cold-outbound mistake cannot touch the domain the company's replies and invoices leave from. SealDeal does not require it and does not sell one. If you use one, connect a mailbox on it and the ramp starts at 20 a day on that domain, separately.
It ramps your volume, automatically, from the day the mailbox is connected, and steps the ramp down when the last week's bounces or complaints say it should. It does not run a warmup network of seeded inboxes exchanging fake mail, and it will not: the mail that leaves is mail a person approved, to a prospect they meant to write to.
The Outbox shows bounce rate, complaint rate, suppressed addresses and whether a hold is active, using the same thresholds the holds enforce, so the number you see is the number the send path acts on. The Operations page shows the hold and the button that releases it.
Related: How the holdout works · vs 11x · Integrations · Security and compliance