Engagement-based filtering is the practice of deciding where your mail lands partly on how recipients have behaved towards it: whether they open, reply, delete without reading, move it out of spam, or report it. It sits on top of authentication rather than replacing it. Passing SPF, DKIM and DMARC establishes who you are; engagement decides what being you is worth.
What signals are actually being used?
Nobody publishes the full set, and treating that as a mystery is a mistake, because the parts providers do publish are consistent and specific.
Google's Postmaster Tools includes a diagnostic that names its own triggers, and two of them are engagement:
- You send too little email.
- Your messages fail to deliver.
- You exceed the recommended spam threshold.
- Recipients indicate that they didn't open or interact with your email.
- Recipients indicate that they want more of your messages.
- You don’t meet our email sender guidelines.
The fourth and fifth entries are the interesting pair. Non-interaction is a negative signal, and wanting more of your mail is a positive one, so the system is scoring in both directions rather than only penalising.
Google's subscription guidance goes further and recommends acting on it: "Consider unsubscribing recipients who don’t open or read your messages." That is a mailbox provider telling senders to shrink their lists, which is only rational if list size with no engagement is actively harmful.
Why is this per-recipient rather than per-sender?
Because the provider knows things about the individual recipient that no aggregate can capture.
A provider can see that this particular user has opened your last twenty messages, or has deleted them unread since spring, or once dragged one out of the spam folder. That history is stronger evidence than your domain-wide reputation, so it can override it in either direction.
The practical consequence surprises people who are used to thinking in campaign averages: your mail does not have a placement, it has a distribution of placements. A campaign can be in the inbox for your engaged core and in spam for the segment you re-added last month, at the same provider, on the same send. An aggregate delivery figure hides that completely, which is one reason seed-based placement measurement exists.
Are opens and clicks the same thing as engagement?
No, and this is where sender-side measurement quietly diverges from what the provider sees.
Your open rate is derived from a tracking pixel being fetched, and plenty of things fetch it that are not people. Spamhaus states its own position on this bluntly:
Please note – Spamhaus reserves the right to engage with the spam we receive and even to follow non-subscription links. There are many anti-spam appliances that now do this also. The fact that an email is opened or a link is clicked does not mean consent has been given.
Privacy features add a second layer of noise by prefetching images for messages that were never opened. So the number on your dashboard and the signal at the provider are related but not the same, and the gap is largest exactly where you most want precision.
Read your own engagement metrics as a directional trend rather than a measurement. If opens fall by half in a week, something happened. If opens are 21 percent, that number is not a fact about human behaviour.
What should I do about an unengaged segment?
Stop mailing it, on a schedule you set in advance rather than in a panic.
The technique is a sunset policy: reduce frequency as engagement drops, ask once, and suppress. Spamhaus publishes a worked example:
- If more than a week passes without the recipient opening an email, change them from weekly to monthly.
- If more than a month passes without the recipient opening an email, send them an email asking, “Do you wish to continue your subscription?” This should contain a link that the contact can click if they wish to continue.
- If they don’t respond, add them to your suppression list and stop sending them emails!
Two reasons to actually do this rather than nod at it. Aggregate engagement is what the provider scores, so carrying a large silent segment drags down the signal for the recipients who do want your mail. And a long-abandoned address is the raw material for a recycled spam trap, so an unengaged segment is not merely inert, it degrades.
List hygiene covers the mechanics of running this on a schedule.
Does sending more fix low engagement?
Almost never, and the belief that it might is why deliverability tends to collapse in seasonal peaks.
The tempting move when results soften is to mail more people more often. Spamhaus has watched this happen for years and describes what senders conclude versus what is actually going on:
Many marketers believe that ISPs, reputation, and filter vendors such as Spamhaus “tighten the rules” around the holidays, which is not the case. Nothing changes on the recipient or vendor side – what causes the perceived “tightening” is created by changes in sender behavior.
The holidays are when senders tend to reach deeply into their subscriber lists (or even buy lists!), taking chances on sending to unengaged or even unsubscribed customers in the hopes of making some extra money during the lucrative holiday sales season.
The filters did not change. The sender reached into a segment that was excluded for a reason, engagement fell, reputation followed, and the result read as an external tightening.
Frequency is worth choosing deliberately rather than maximising. Set the expectation when someone subscribes, then keep to it: as Spamhaus puts it, "people who have a clear expectation that is met will feel respected and are much less likely to report as spam, delete the emails unread, or unsubscribe."