Complaints and list hygiene — Advanced Email Trust

Two numbers determine most of a sending reputation, and both are consequences of how a list was built rather than of anything technical.

Complaint rate

The proportion of delivered messages a recipient marks as spam. Google publishes a threshold, which makes it the one number in deliverability with a public line to stay under.

below 0.10%   fine
0.10% - 0.30% the zone where filtering begins to tighten
above 0.30%   Gmail's stated limit. Expect degradation.

For context: 3 complaints in 1,000 delivered messages
is the limit. Not 3 in 100.

It is measured per sending domain, so one badly run campaign damages your transactional mail if both send as the same domain. This is the strongest argument for separating streams onto subdomains.
A complaint is a recipient action, not a bounce. No amount of infrastructure work reduces it; only sending mail people want does.
Feedback loops let you act on it. Most large providers will report complaints back to a registered sender, and the correct response is immediate suppression of that address.

Invalid addresses

Sending to addresses that do not exist tells a receiver the list was not collected with consent or has not been maintained. A few percent is normal decay; a high rate on a first send indicates a purchased or harvested list, and is treated accordingly.

Suppress a hard bounce permanently, on the first occurrence.
Suppress a soft bounce after a few consecutive failures.
Never re-add a suppressed address because it appeared in a new import.

Spam traps

Addresses that exist only to catch senders. A pristine trap was never a real address and can only reach a list through harvesting or invention. A recycled trap was a real address, abandoned, and repurposed after a long dormancy — so hitting one says you are mailing people who stopped existing years ago and never noticed.

Recycled traps are what sunsetting prevents

A subscriber who has not opened anything in two years is either uninterested or gone. Continuing to mail them earns no engagement and risks a recycled trap. Removing them feels like discarding reach and is the single most effective hygiene measure available.

The practices that work

Confirmed opt-in. More friction, far better lists, and it makes trap hits nearly impossible.
Suppression that is honoured everywhere. Across every platform, permanently.
Sunsetting on engagement. Stop mailing people who have stopped reading.
Separate streams by subdomain. Marketing and transactional mail earn their reputations independently.
Never buy a list. There is no version of this that ends well.

Knowledge check

Marketing and transactional mail both send as example.com. A campaign pushes the complaint rate to 0.4%. What is affected?

Both — reputation attaches to the sending domain, so password resets and receipts are filtered alongside the campaign.
Only the marketing mail, since receivers score per campaign.
Nothing, provided all the mail passes DMARC.
Only mail to the provider that reported the complaints.

Correct — Both — reputation attaches to the sending domain, so password resets and receipts are filtered alongside the campaign. Sharing a domain means sharing a reputation. Separating the streams onto subdomains is what lets a bad campaign damage only the campaign, and it is the most valuable structural decision in bulk sending.

Not this — Only the marketing mail, since receivers score per campaign. Receivers score domains and IPs over time. There is no per-campaign identity for them to score.

Not this — Nothing, provided all the mail passes DMARC. Authentication makes mail eligible to be judged. The complaint rate is the judgement.

Not this — Only mail to the provider that reported the complaints. That provider acts on what it measured, and the underlying list problem produces complaints everywhere.