What are weighted signals in spam filtering?
Still have a question, spotted an error, or have a better explanation or a source we should cite?
Think of it like a courtroom. Not every witness carries the same weight. A missing DKIM signature might nudge the score up a little. A sending IP that's on a blocklist might slam it up a lot. That's the whole idea behind weighted signals in spam filtering.
Weighted signals are the individual factors a spam filter checks, each assigned a point value based on how strongly that factor predicts spam. Together, those points build your spam score. Cross the threshold, and you're filtered.
The weights aren't arbitrary. They come from analyzing massive volumes of real spam. If a particular pattern shows up in 90% of spam campaigns but almost never in legitimate mail, it gets a high weight. If a pattern appears in both spam and clean email, it gets a low weight (or gets ignored entirely).
A few examples of how the math actually works:
- Missing DKIM signature: roughly +1 point
- Sending from a blocklisted IP: +4 points or more
- Matching a known phishing pattern: +5 points or more
- Spam-trigger words in the subject: +0.5 to +2 points depending on the word
The highest-weighted signals are the ones that almost exclusively appear in malicious or low-quality mail. That's why authentication failures, blocklist hits, and broken sending infrastructure hurt so much more than a single awkward subject line.
Server administrators can also customize weights for their own environment. If a rule keeps flagging legitimate emails as spam (a false positive), they can reduce that rule's weight or turn it off entirely. So a filter at a medical company might lower the weight on drug-related terms that a general filter would treat as red flags.
The practical takeaway: fix the big stuff first. Strong authentication (SPF, DKIM, DMARC), a clean sending IP, and a healthy list will pull your score down more than perfecting your subject line ever will. If you want to see what signals your email is actually triggering, our free Email Header Analyzer is a good place to start.
The most useful email questions, once a month
One email a month. New questions worth reading, real deliverability changes (Gmail and Outlook updates, etc.), and any tool or report that landed since the last issue. No fluff, no upsell.
Contributors
Who worked on this answer
Every name links to their profile. Every company links to their site. Real people, real accountability.
