How are feedback loops integrated into filter scoring?

Feedback loops (FBLs) are the pipes mailbox providers use to tell you when their users hit the spam button on your mail. You sign up, you get a stream of complaint reports, and the same complaints that land in your inbox also feed the provider's internal reputation systems. So FBL data shapes filtering twice: once because you act on it, and once because the provider already counted it against you.

The big four feeds you should know:

• Microsoft JMRP (Junk Mail Reporting Program). Outlook.com, Hotmail, Live. You receive per-message ARF (Abuse Reporting Format, the standard complaint format defined in RFC 5965) reports for every user who clicks Junk. Since June 2026 those reports hide who clicked, so match them back through your own Message-ID or a custom X-header. (Sources: Spam Resource and emailexpert.)
• Microsoft SNDS (Smart Network Data Services). Not a per-message feed. SNDS shows IP-level data: complaint rate buckets, spam trap hits, filter result, RCPT counts. It is how Microsoft tells you what your IP looks like to them.
• Yahoo Complaint Feedback Loop. Per-message ARF for Yahoo, AOL, and the rest of the Yahoo Mail family. You register the DKIM d= domain rather than the IP.
• Google Postmaster Tools. Google does not give you per-message complaints. They publish aggregate domain reputation, IP reputation, spam rate, authentication pass rates, and delivery errors. The spam rate dashboard is the closest thing you get to a Gmail FBL. See Postmaster Tools docs.

A few smaller providers (Comcast, La Poste, Mail.ru, Fastmail via AOL) run their own ARF feeds. M3AAWG keeps a working list and best practice docs worth reading.

How this actually moves filter scores. One complaint does nothing. Providers care about rate, not raw count, and they aggregate over rolling windows. The number everyone quotes is 0.1% complaints (1 per 1,000 delivered) as a soft ceiling, 0.3% as the line where Gmail and Microsoft start dropping you to spam and throttling. Postmaster Tools labels anything above 0.3% as red. Yahoo will quietly start junking before they ever block you.

What to do with the data when it lands:

1. Suppress the complainer the same day the report arrives. Mailing someone who already clicked Junk is the fastest way to make the next complaint worse. Your ESP can usually auto-suppress Yahoo CFL hits if you wire the webhook. JMRP hits only suppress automatically if your messages carry an ID you can match back to the recipient.
2. Group complaints by campaign, segment, and acquisition source. If one signup form is producing 80% of your complaints, the form is the problem, not the send.
3. Watch SNDS and Postmaster Tools weekly, not just when something breaks. Reputation moves slowly. By the time you are blocked, the chart has been yellow for a month.
4. Cross-check complaint spikes against content changes. A new From name, a re-engagement subject line, or an old list segment can all push the rate up in a single send.

FBLs do not replace ongoing sender behavior monitoring or the other signals filters analyze, but they are the most direct line you get into how recipients feel about your mail. If you ignore them, the filter will tell you what the complainers already said, just six weeks later and at the cost of your inbox placement.