List hygiene: why a smaller engaged list delivers better

Every mailbox provider keeps its own private view of whether mail from your domain is wanted. No provider publishes the full formula, and the reasonable assumption is that it differs between them, so any article claiming to list the weights is guessing. What is documented is which signals feed it: how often recipients mark your mail as spam, whether they engage with it, how many addresses bounce, whether you hit spam traps, whether your mail authenticates, and whether your volume is steady or arrives in unexplained spikes.

One number in that set is published. Google's sender guidelines tell senders to keep the spam rate reported in Postmaster Tools below 0.30%, and the practical advice from the same guidance is to stay under 0.10%. That is a genuinely useful figure precisely because it comes from the receiver rather than from a vendor's estimate.

Reputation is also attached to a domain and a sending IP rather than to a campaign. This is why one bad send is rarely catastrophic and why a slow drift is rarely noticed until delivery is already poor.

Filtering is comparative. A provider decides where to put your mail partly by how recipients of your previous mail behaved, so a batch sent largely to people who never interact produces very little positive signal and a proportionally larger share of negative signal from the few who do react — by complaining.

There is a harder mechanism underneath. Abandoned mailboxes do not stay abandoned forever; providers recycle some of them into spam traps, addresses that exist only to catch senders mailing lists they have not maintained. Hitting one is not a soft warning. It is evidence that you are mailing addresses with no consent behind them, because a real subscriber would have noticed the mailbox was gone.

And every message costs something to send, whether that is priced per message or per subscriber. A list padded with addresses that will never buy anything is a line item that grows quietly while the return on it stays at zero.

The obvious method — find everyone who has not opened in months — has stopped working reliably. Apple Mail's Privacy Protection routes messages through Apple's own proxy and preloads images, tracking pixels included, before the recipient has done anything. An address that has never been looked at can therefore report opens indefinitely.

So build the definition on actions that require a human: clicks, replies, purchases, logins, bookings, site visits attributable to the email. These are fewer and noisier than opens, and they are real.

Then choose a window that matches how your business actually works rather than importing a generic ninety-day rule. A content newsletter with a weekly cadence can reasonably treat ninety days of silence as dormant. A Kerala homestay whose guests book once a year has no business calling somebody inactive at six months. The window is a statement about your purchase cycle, and getting it wrong in the strict direction throws away customers who were behaving normally.

Before removing anyone, ask them. A short series — two or three messages over a couple of weeks — that says plainly that you have noticed they have not been reading, explains what they would keep receiving, and offers an obvious way out.

The honest version does not disguise the question. A win-back message built around a discount tests whether people want a discount, not whether they want your email, and it teaches the ones who stay to wait for the next one. Asking directly, with a single clear confirmation link, produces a smaller list and a truer one.

Two operational cautions. Do not send the whole dormant segment at once: it is your least engaged mail, so a large simultaneous batch is the worst possible signal to send in one burst. And send it from your established sending setup rather than a fresh domain, since a new domain has no reputation to spend on a difficult audience.

When someone is removed, keep a record of the removal. Deleting the address entirely feels like cleaning, but it destroys the only evidence that they ever asked to leave — and it guarantees they come back. The next time somebody imports a spreadsheet exported before the cleanup, every deleted address is re-subscribed, and the person who unsubscribed six months ago receives your mail again.

A suppression list is the fix: addresses your system holds specifically in order never to mail them. It is memory rather than a mailing list, and it should outlive any individual platform you use. When you migrate providers, the suppression list migrates with the subscriber list or the migration is incomplete.

This matters beyond tidiness. If a recipient ever disputes that they opted out, the suppression record and its timestamp are what let you show the request was honoured, and when.

Most hygiene problems are acquisition problems arriving late. A list assembled from purchased data, conference badge scans or a competition entry is full of addresses that never asked for anything, and no amount of pruning later converts them into subscribers.

Confirmed opt-in — where the address has to click a link before it receives anything else — is the single strongest control. It costs you some signups, and every address it removes is one that would have damaged reputation later, including the typos and the deliberately fake entries.

A few smaller measures are worth the effort: catch common domain typos at the form rather than mailing them and taking the bounce; use a hidden honeypot field to absorb automated submissions; and state clearly at the point of signup what you will send and roughly how often. That last one is not a legal formality. It is the expectation the subscriber will judge you against, and a subscriber who got what they expected is the one who does not report you.