Check this number before you check your reply rate

Most founders check reply rate every week and domain health never, until the week reply rate crashes and there’s no obvious reason why. By then you’re debugging backward — reading DMARC reports, checking blocklists, trying to figure out what changed — when the actual answer was visible days earlier, in a number you weren’t in the habit of looking at.

Reply rate tells you how last week’s campaign performed. Domain health tells you whether next week’s campaign is going to reach anyone at all. Those are different questions, and the second one is the one worth checking first.

What actually feeds the score

A domain health score rolls up several separate signals mailbox providers track individually into one number, so you’re not cross-referencing five dashboards to answer “is this domain okay”:

Bounce rate — how often mail sent from this domain fails to deliver. A rising bounce rate is usually the earliest signal something’s wrong, whether that’s list decay from an aging import, a typo-riddled manual add, or a sudden authentication problem that’s causing valid addresses to reject mail they’d normally accept. It’s also one of the fastest signals to fix once caught: a stale segment of a list can be re-verified or excluded well before it does lasting damage to reputation.

Spam complaint rate — how often recipients mark mail from this domain as spam. This is the single most consequential signal to mailbox providers, because a complaint is a person actively telling the provider they didn’t want the mail — a much stronger negative signal than simply not opening it. A handful of complaints on a small send can move this number more than most senders expect, which is part of why the enforced threshold sits as low as it does.

Authentication pass rate — how consistently mail from the domain passes SPF, DKIM, and DMARC alignment. Authentication failures don’t just risk one message; they undermine the trust signal every other message from the domain depends on, since a provider that can’t verify a sender is who it claims to be has no reliable basis for treating any of its mail as legitimate.

Sending consistency — whether volume ramps predictably or spikes erratically. A domain that jumps from a trickle to a flood in one day reads as suspicious to a mailbox provider even if every individual message is legitimate, because that pattern is also what a compromised account sending spam typically looks like from the outside.

Reply engagement — positive signal from the people actually receiving the mail: opens, clicks, and replies that indicate the domain is sending something recipients want, not just something that technically delivers. Engagement is the counterweight to the other four signals, which are all measuring absence of harm — this one measures actual, positive interest.

Why the thresholds are tighter than most senders assume

The instinct is to treat “under some vague spam number” as good enough. Google’s own guidance is more specific than that, and it tightened in 2026: a spam complaint rate that mailbox providers used to flag around 0.3% is now enforced closer to 0.1% before Gmail’s Postmaster Tools marks a domain’s deliverability status as a problem. That’s not a soft recommendation — it’s the documented threshold behind the plain-language verdict Google now shows directly in its Postmaster dashboard: whether recipients actually want your mail.

Why one number beats five dashboards

The value of rolling these signals into a single score isn’t precision — it’s speed of decision. A founder checking analytics once a week doesn’t have time to read a bounce-rate trend, a spam-complaint percentage, an authentication pass rate, and a send-volume curve separately and mentally combine them into “is this domain okay to keep sending from.” A single score, watched over time, answers that faster than the five inputs it’s built from ever could on their own — and it’s the same principle behind what a DMARC report is actually telling you: raw signal is only useful once someone’s actually translated it into a decision.

This is also why the score matters more than any single input read in isolation. A bounce rate that ticks up slightly might be nothing — a handful of role addresses that changed hands. A spam complaint rate that ticks up slightly might also be nothing — a naturally noisier week on a larger send. But two or three of those signals drifting the wrong way at the same time is a genuinely different situation than any one of them moving alone, and that’s exactly the pattern a rolled-up score is built to catch that a person skimming five separate charts is likely to miss.

What happens automatically when it drops

The score isn’t just a read-only number to glance at — it drives an actual response. When a mailbox’s health signals start trending the wrong way, sending volume on that specific mailbox backs off automatically rather than continuing at full pace into a worsening problem. That protects the domain the way a smoke detector protects a building: not by preventing every fire, but by catching the early signal before it becomes the kind of damage that takes weeks to recover from.

That automatic back-off matters more than a dashboard alert by itself, because the failure mode with alerts alone is well known — a warning sits unread in an inbox for two days while volume keeps pushing into a declining trend. A score that actually changes sending behavior closes that gap without depending on someone checking a dashboard at exactly the right moment.

What a dip actually looks like in practice

The realistic version of this isn’t a score that falls off a cliff overnight — it’s a slow drift that’s easy to miss if nobody’s watching. A mailbox’s bounce rate creeps up over ten days because a chunk of an imported list has gone stale. Spam complaints tick up slightly the same week, because a subject line that tested fine on a smaller list starts reading as spammier at higher volume. Neither change is dramatic enough to trigger obvious alarm on its own — the campaign is still sending, opens are still happening, nothing looks broken. That’s exactly the window where a rolled-up score earns its keep: two small negative signals moving together pull the number down before either one alone would have been enough to make you look twice.

By the time reply rate visibly drops — the metric most founders actually check — the underlying reputation damage has usually been building for a week or two already. Domain health exists to surface that drift while it’s still a ten-point dip and a sending back-off, not after it’s become a blocklisting event that takes a month to recover from.

One score per mailbox, not one score for the whole account

A domain sending through several connected mailboxes doesn’t get a single blended number that hides which address is actually the problem. Each mailbox carries its own health signal, because reputation is tracked per sending address by mailbox providers, not per company. That granularity is what makes the automatic response useful instead of blunt: if one of five connected mailboxes starts trending badly, only that mailbox’s sending volume backs off, while the other four keep carrying the campaign at full pace. A single account-wide score would either overreact — throttling healthy mailboxes because one is struggling — or underreact, hiding one bad mailbox inside an average that still looks fine.

What “healthy” actually protects you from

The worst-case outcome domain health monitoring exists to catch isn’t a slightly lower open rate — it’s a blocklisting event, where a mailbox provider stops accepting mail from the domain outright. Recovering from that is measured in weeks, not days, and it affects every campaign running on that domain, not just the one that triggered it. Google’s Postmaster Tools changes in 2026 made the warning signs more explicit than they used to be — the platform now tells you directly whether recipients want your mail — but that only helps if something’s actually watching the signal and acting on it before a human notices the campaign has gone quiet.

Getting started

Domain health monitoring runs continuously on every connected sender, included in DMARC monitoring on every Norbelys plan from Starter up — it’s not a metric you have to remember to check manually, and it’s not something you upgrade into on a higher tier the way some deliverability tooling gates advanced monitoring behind a premium plan. Start on any plan, connect your sending domain, and let the score do the job most founders currently do by hand: noticing their reply rate crashed, a week too late to have done anything about the cause.

Domain health, answered directly

What's the difference between domain health and reply rate?+

Reply rate measures how a specific campaign performed after it sent. Domain health measures whether the domain is currently trusted enough by mailbox providers to reliably reach the inbox at all — it's a leading indicator, where reply rate is a trailing one.

What actually causes the score to drop?+

Rising bounce rate, rising spam complaints, authentication failures (SPF, DKIM, DMARC), erratic sending volume, and declining reply engagement are the core inputs — any of them trending the wrong way pulls the score down.

Does anything happen automatically if the score gets bad, or do I have to react manually?+

Sending volume on the affected mailbox backs off automatically when health signals trend badly, rather than continuing at full pace into a worsening problem — the response doesn't depend on someone catching a dashboard alert in time.

Is this the same thing as a DMARC report?+

Related but not identical. DMARC monitoring reports on authentication results specifically; domain health rolls authentication together with bounce rate, spam complaints, sending consistency, and engagement into one score that reflects overall sender reputation, not just authentication alignment.

If I connect several mailboxes on one domain, do they share a single score?+

No. Each connected mailbox has its own health signal, tracked independently, because mailbox providers track reputation per sending address rather than per company. That means an issue on one mailbox triggers a volume back-off on that mailbox specifically, without throttling the others.

How quickly does the score reflect a change, good or bad?+

The underlying signals — bounce rate, spam complaints, authentication results, engagement — update continuously as mail sends and responses come in, so the score reflects recent sending behavior rather than a snapshot from whenever you last happened to check it.