Reputation recovery: the 90-day playbook for dropping from High to Low

Here's the mental model. Reputation is a moving average of how recipients have been treating your sends over a rolling window. Spam complaints, bounces, and being marked as junk are the negatives. Opens, clicks, and being moved out of spam are the positives. The window doesn't care that you've had a heart-to-heart with the team and resolved to do better. It only updates as fresh sends produce fresh signal.

Which means the conversation with stakeholders has to happen on day one. Six weeks minimum before Postmaster Tools starts drifting back toward Medium. Twelve weeks to reach High. Push volume during recovery and you extend the timeline — recovery moves exactly as fast as your willingness to throttle, and not a day faster.

The sooner you say that out loud, the better the rest of the quarter goes. A finance team told "two weeks" in week one gets angry in week six. A finance team told "twelve weeks" in week one is pleasantly surprised in week nine. Same outcome, two completely different political situations.

Before anything else, find what caused this. Was it a complaint spike from a dormant segment getting reactivated into a campaign? A new ESP — email service provider, the platform you send from — or a new sending IP? An authentication regression where SPF, DKIM, or DMARC quietly stopped passing? The fix depends entirely on the root cause, and skipping the diagnosis to "move fast" is how teams spend week six discovering they rebuilt the wrong thing.

Three moves, in order:

Suppress dormants immediately. Anyone whose last open or click is more than 30 days old comes off marketing today. This is the single biggest lever in week one — it cuts the bad-signal volume at source and lets the rolling window start to see clean sends.

Audit authentication. SPF, DKIM, and DMARC are the three protocols that prove an email actually came from your domain — together they tell mailbox providers "this isn't a forgery." All three need to be passing at 99%+ in Postmaster Tools. Any dip gets fixed here first. Recovering reputation on unauthenticated mail is impossible — the receiving server can't tell your recovery sends from someone spoofing you.

Audit recent content. Subject lines pinging spam filters? Image-heavy templates with barely any text? Shortened URLs from a sketchy domain? Fix the obvious ones. You're not hunting for perfection — you're removing the things that made last week's numbers worse.

Imagine the reputation algorithm as a wary doorman. Three weeks ago you turned up with a rowdy crowd and got barred. Now you're back, and every guest you bring with you is being scrutinised. The job for the next three weeks is to only bring guests who'll behave.

Engaged users plus known-good content plus reduced volume equals the signal mix that repairs reputation. That's the entire formula. The discipline is in not deviating from it.

Audience: 30-day engaged only. Anyone whose most recent open or click sits inside a 30-day window from today. This is your safe cohort — high probability of positive signal per send, near-zero probability of complaint.

Volume: 30–50% of pre-incident. If you were pushing 500K/day, you're pushing 150–250K/day now. Ramp slowly or not at all.

Content: your best-performing campaigns only. Not promotional-heavy broadcasts. Not the experimental template you've been itching to test. Your known-winners, reused.

Monitoring: daily. Postmaster Tools, complaint rate, bounce rate. Any regression and you throttle further. Discipline here is boring and load-bearing — the temptation will be to push slightly more "because it looked fine yesterday." Don't.

If Postmaster has drifted from Low to Medium, and the complaint rate is sitting below 0.2% (two complaints per thousand sends), the doorman is starting to relax. You can begin expanding. Cautiously.

Engagement window: 30 days becomes 60 days. You're bringing in slightly less-engaged users, but only after the recent baseline has stabilised.

Volume: 60–75% of pre-incident.

Content variety: seasonal campaigns and launches come back in — but as smaller test sends first, not to the full reactivated file.

90-day dormants stay suppressed. These are the users who triggered the original incident. Keep them off.

If Postmaster hasn't moved by week five, the recovery isn't working. Almost always because the root cause wasn't fully addressed in week one. Re-diagnose before pushing further volume — compounding a broken send into a broken reputation makes week eight worse, not better. The cheapest re-diagnosis is to assume you missed something and look harder.

Postmaster at Medium, complaint rate below 0.15%, bounces stable. You're through it.

Volume back to 100%. Full pre-incident restored.

Audience expansion. 90-day engagement window; some dormants re-enter via win-back flows — automated sequences designed to re-engage cold users without dumping them straight back into the broadcast list (winback flows guide). Never re-include the full dormant cohort at once. That's how incidents recur.

Permanent list hygiene. The hygiene policy becomes standing operations, not a one-time cleanup. If the policy holds, the next incident doesn't happen.

The hardest part of recovery isn't the sending plan. It's the meeting where you tell finance the program will run at half-volume for three months. Here's a workable week-one comms template for executives:

"We've identified a deliverability issue — our sender reputation with Gmail has dropped. The cause is [X]. Recovery is expected to take 6–12 weeks. During recovery, sending volume reduces to 30–50% of normal, focused on our most engaged users. Revenue impact during recovery: ~$[estimate] below forecast. Once recovered, we'll have a revised list-hygiene policy that prevents recurrence."

Specific, honest, dated. Avoid the trap of soft-pedalling the timeline to make the news easier — a two-week promise that becomes ten is far worse politically than a ten-week promise up front. The finance team can model a long recovery. They cannot model a lie.

A few questions that come up in every recovery, with the short answers:

Can we keep sending transactional? Yes — and you should, if transactional sits on a separate subdomain (something like notifications.yourbrand.com) with its own reputation. If it doesn't, transactional is currently riding the damaged marketing domain, which is a great argument for setting up subdomain separation as part of this project.

Should we switch IPs? No. Modern ISPs weight domain reputation over IP reputation, so a new IP resets nothing meaningful while adding an IP-warmup project — the slow ramp every new IP needs to establish itself with mailbox providers — on top of the recovery you're already doing. Stay put.

What's the revenue hit? Roughly proportional to the volume cut. Sending 40% of normal to 40% of the list lands you at 30–50% of normal revenue in weeks 2–4, 60–80% in weeks 5–8, back to normal by week 12. Model it into the comms so finance isn't surprised.

The Deliverability Management skillcovers this recovery sequence as one of its core playbooks. Having it documented before you need it is the difference between a structured recovery and a panicked one — and panicked recoveries are how programs end up in week fourteen still explaining to the CFO why this isn't fixed yet.