The unglamorous work behind good email deliverability

When open rates drop, the first instinct is always creative: rewrite the subject line, test the send time, redesign the template. It's the visible layer, so it gets the attention. But if your emails aren't reaching inboxes in the first place, none of it matters - and whether they reach inboxes is decided by things nobody puts in a portfolio: authentication records, ramp schedules, suppression lists, and the discipline to stop mailing people who stopped reading.

The stakes got concrete recently. Gmail, Yahoo, and now Microsoft all enforce hard requirements on bulk senders: anyone sending 5,000+ emails a day needs SPF, DKIM, and DMARC in place, a spam complaint rate under 0.3%, and one-click unsubscribe. And enforcement has teeth now - Google moved from temporarily delaying non-compliant mail to permanently rejecting it in late 2025, and Microsoft began rejecting non-compliant bulk mail in May 2025. Deliverability stopped being a dark art and became a compliance floor with published thresholds.

Here's the actual work, in order of how often we find it missing.

The floor: authentication and the 0.3% line

The technical prerequisites are now non-negotiable: SPF and DKIM aligned, a DMARC policy published, a working one-click unsubscribe honoured within two business days. This is a day of DNS work, and it's table stakes rather than an advantage.

The number that deserves more respect is the complaint threshold. The 0.3% spam complaint rate is where enforcement begins, but Google's own recommendation is to stay below 0.1% for reliable inbox placement - the gap between those two figures is your margin of safety, and it's thin. At 0.1%, one complaint per thousand emails. A single careless send to a cold segment can eat a month of good behaviour, which is exactly why the rest of this article is about who you mail, not what you write.

Warm-up: reputation is earned on a schedule

Mailbox providers treat volume from an unknown domain or IP the way a bank treats a first-time borrower: no history means no trust. The fix is boring by design. Warming up means methodically adding volume over several weeks so providers can watch how recipients react - and the reaction is the point, which is why the first weeks go exclusively to your most engaged recipients, people who opened or clicked in the last 30 days, whose opens and replies tell the algorithm this sender is wanted.

The shape of a plan that works: start small, double weekly rather than daily, and never jump volume by more than about 20% in a single day, widening the audience in rings - 30-day engagers first, then 60-day, keeping anyone past 90 days out of the first six weeks entirely. Expect four to eight weeks to reach full volume, and expect consistency to matter as much as the numbers - spiky, stop-start sending reads as suspicious even at low volume.

The same logic applies beyond new domains: after a long pause, after a platform migration, after a rebrand onto a new sending domain. Reputation doesn't transfer; it re-accrues.

List hygiene: where most problems actually start

Here's the uncomfortable arithmetic. Email databases decay at roughly 23% per year - people change jobs, domains get retired, inboxes get abandoned. A list you built three years ago and never pruned is now a minefield of dead addresses, and some of those dead addresses have been recycled by mailbox providers into spam traps: addresses whose only purpose is to catch senders who don't clean their lists.

So "we email our whole database" is not a strategy, it's a slow-motion incident. Every send to an invalid address is a bounce that dents your reputation; every send to an abandoned inbox is a non-open that drags your engagement metrics down; and the occasional trap hit can hurt far more. The correlation is stark: senders with hygiene scores above 95% see around 97% inbox placement, while those below 85% manage just 76%. That 20-point gap is bigger than any subject-line test will ever deliver.

The hygiene routine is simple and relentless: validate new addresses at the point of capture, run periodic verification on the existing base, remove hard bounces immediately and automatically, and treat repeated soft bounces as hard after a few attempts. None of it is clever. All of it compounds.

Suppression discipline: the sunset policy

This is the part teams resist, because it means deliberately mailing fewer people. A sunset policy defines when an unengaged subscriber stops receiving regular sends - and fewer than a quarter of senders have one, which goes a long way to explaining the general state of B2B inboxes.

The mechanics: define inactivity in terms of your cadence - common windows run 90-180 days without an open or click for frequent senders, stretching to 9-12 months for lighter programmes. When a contact crosses the line, they get a short win-back sequence - one genuine "do you still want this?" attempt, not five. No response means suppression: the contact stays in your CRM with full history intact, but is excluded from regular sends. Suppression is not deletion - the record, the consent trail, and the history all survive. What stops is the damage.

The objection is always the same: "but they might convert eventually." The honest answer is that someone who hasn't opened an email in a year is not converting from email - but they are suppressing the metrics that mailbox providers use to judge every email you send to everyone else. Keeping them in the audience taxes deliverability to your best subscribers. That's the trade, and stated plainly, nobody takes it.

Engagement scoring: segmentation as a reputation tool

Between "fully engaged" and "sunset" sits everyone else, and this is where engagement scoring earns its keep - not as a vanity dashboard, but as send logic. A simple recency-based tiering works: highly engaged (opened or clicked in the last 30-60 days) can receive your full programme; moderately engaged get reduced frequency and your strongest content only; lapsing contacts get the win-back treatment; suppressed contacts get nothing without explicit re-permission.

The effect is that your send volume concentrates where the positive signals are - which is exactly the population that makes your complaint rate low and your engagement rate high. Blast-everything sending does the opposite: it maximises exposure to the people most likely to ignore or report you. Unsegmented sends erode reputation faster than any single bad campaign, because they repeat the same small damage weekly until the damage is structural.

The compounding effect

None of this is a project with a launch date. Authentication is set once and monitored; warm-up happens when infrastructure changes; hygiene, suppression, and scoring run continuously or not at all. That's precisely why it works as a competitive advantage - most senders won't sustain it, because there's nothing to show in a Monday meeting except a complaint rate that stayed boring and an inbox placement number that didn't move.

But that's the win. Deliverability isn't a lever you pull before a big campaign; it's the accumulated interest on months of unglamorous decisions about who not to email. Get that right, and the subject-line tests finally have an audience to land in front of.