At some point you did the responsible thing.
You stopped blasting the whole list.
You built a segment… recently engaged, active, however your ESP words it, and you started sending to that instead.
Placement improved.
You felt like an operator instead of a spammer.
That segment is almost certainly built on opens.
And opens are the one signal that stopped meaning anything.
This one's for anyone sending daily or near-daily who has already built an engaged segment and assumed it was working.
Three things in ninety seconds:
The segment you call engaged is probably an open-based segment
Opens over-count the dead and under-count the living, at the same time
Cleaning the list harder won't fix it, you're missing a control lever, not hygiene
The Ten-Minute Check
Open your ESP.
Pull your engaged segment, whatever definition you're currently sending to.
Now filter it down to subscribers who have never clicked anything. Not even once.
Divide that count by the size of the segment.
That's your ghost share, and if you've never run this, the number is going to be higher than you're expecting.
Those subscribers qualified as engaged because a pixel fired.
Nobody was there.
Apple Mail Privacy Protection loaded the image before the message was ever opened, or an image prefetcher pulled it, or a corporate security scanner walked every asset in the email including the tracking pixel.
They passed your engagement filter without a human being involved at any point.
If your ESP has an MCP connection, you can skip the interface:
For [Newsletter Name], how many subscribers in my engaged segment have never recorded a click? Give me the count and the percentage of the segment.
Why This Is Worse Than It Looks
Most people hear this and think they have a counting problem.
A little inflation, no real harm.
It runs in both directions at once, and that's what makes it dangerous.
Over-counting keeps the dead.
Ghosts stay in your sending pool indefinitely.
They never click, never convert, never complain… they just quietly sit there absorbing volume and dragging the engagement ratio the ISPs are actually measuring.
Under-counting removes the living.
Subscribers reading with images disabled register zero opens no matter how engaged they are.
Some of them click.
Your open-based filter flags them inactive and sunsets them anyway.
So you're keeping the subscribers who do nothing and cutting the ones who do something.
That's not a slightly noisy segment.
That's a segment doing the opposite of its job.
And you can't tighten your way out of it.
A ghost fires a pixel on every single send, it is the most reliable opener on your list.
It survives any window you set.
Narrow the window and the irregular humans fall out first, while every ghost stays exactly where it is.
The window length was never the broken part.
The Part That Actually Matters
Here's what I want you to sit with.
When your placement slips, you have exactly two moves available right now.
Send less. Or clean harder.
Both are reactive. Both are blunt.
Neither responds to what the ISPs are telling you, because you have no mechanism that reads those signals and adjusts.
That mechanism is what a Base Sending Segment is.
Not a segment you build once and forget.
A dial, one you tighten when Postmaster signals turn, and loosen when they recover.
Built on clicks, because clicks are the only engagement signal a machine doesn’t fake at scale on a subscriber's behalf on the top inbox provider Gmail.
And the windows on that dial don't count calendar days.
They count sends.
A thirty-day window means something completely different at twice a day than it does at once a week, and almost everyone gets this backwards.
I'm walking through the whole thing, the branch logic, the actual window defaults I'm running right now, the cadence table, and what happens to subscribers who fall out, at the BSS workshop on Thursday, September 17th.
Run your ghost number first.
The workshop makes considerably more sense when you already know how bad yours is.
What To Do Before Then
You don't have to wait for me to start.
Stop treating an open as evidence a human was present.
Anywhere in your stack where an open alone qualifies someone as engaged, that rule is now suspect.
Pull your ghost number and write it down.
It's your baseline, and you'll want it when you rebuild.
And don't expect a shorter open window to solve this on its own.
Tighten all you like, ghosts open everything, every time, and they'll clear any bar you set.
What has to change is that an open is qualifying someone by itself.
Your engaged segment was a good instinct.
It's just measuring the wrong thing, and it has been since Apple shipped Mail Privacy Protection five years ago.
Go count your ghosts,
Chris Miquel
P.S. If your ghost share comes back ugly, don't try to fix it this week.
Knowing your segment is broken and knowing what to replace it with are two different problems. Guessing at the second one is how operators end up cutting their sending list in half and getting nothing for it.
Bring your number to the workshop on Thursday, September 7th, and we'll build the replacement together — with the actual windows, not the theory.
