Feedback loops: getting complaints back, and acting on them

When someone marks your message as spam, the provider knows. Whether you find out depends on whether you registered for a feedback loop.

Without one, complaints are invisible: delivery quietly degrades and the cause is not in your logs. With one, each complaint arrives as a message you can attribute and act on.

Who offers what

Microsoft (JMRP) — the Junk Mail Reporting Program. Register your sending IPs, prove you control them, and complaints arrive as ARF messages. Pair it with SNDS, which gives complaint rates and trap hits for the same IPs. This is the most useful of the consumer feedback loops.

Yahoo and the Verizon Media properties — a complaint feed tied to a registered domain and DKIM selector rather than to IPs.

Google — no per-complaint feed. The FBL is aggregate and identifier-based: you attach a Feedback-ID header to outgoing mail and Postmaster Tools reports complaint rates per identifier. You learn which stream is generating complaints, never which recipient.

Smaller providers — many participate through ARF. Worth registering with the ones your list actually contains, which you can find by counting recipient domains.

The Feedback-ID header

Google's scheme is worth setting up properly because it is the only visibility you get there. The header has up to four colon-separated fields, with the sender identifier last:

Feedback-ID: campaign123:newsletter:acct456:yourdomain

Choose fields that are stable and meaningful — a stream, a campaign, a customer account. Identifiers that change every send produce reports with sample sizes too small to mean anything.

Acting on what arrives

A complaint is a permanent suppression. Immediately, across every stream. There is no threshold and no second chance: the recipient has told their provider your mail is junk, and the provider is now applying that to everything from you.

Automate it. A complaint that sits in a mailbox until someone reads it on Monday is a weekend of sending to someone who already complained.

Reading the trend

The individual complaint matters less than the rate and its shape.

• A spike after one send — something about that send. Wrong list, wrong content, a subject line that misled.
• A steady climb — list quality degrading, or sending frequency that has crept up past what recipients signed up for.
• Concentrated in one provider — often a filtering change rather than a behaviour change. Compare against your own delivery data before rewriting everything.

What good looks like

Below 0.1%. The stated ceilings are higher — Gmail cites 0.3%, Microsoft is stricter at under 0.2% — but those are the lines at which action is taken, not targets. A sender at 0.1% has room to absorb a mistake. A sender at 0.25% does not.

See reducing complaint rate for what to do when the number is already too high.