Everyone gets told the same thing about deliverability. Stop sending to everyone, build an engaged segment, send there instead.
It is correct, and it is the single highest-value change most stores can make to their email. Then it gets implemented as "opened in the last 30 days", because that is the number in the article, and a store that emails twice a month cuts its addressable list by more than it needed to and cannot work out why revenue dropped.
The window is not a constant. It is a function of how often you send, and getting it wrong in the tight direction is expensive in a way that is difficult to see.
Why the window depends on your cadence
An engaged segment asks: has this person shown a sign of life recently?
The only way they can show one is if you gave them the chance. Someone who receives eight emails a month has eight chances to open one inside a 30-day window. Someone who receives two has two.
Both of those people can be equally interested in your store. On a 30-day window, the second one falls out of your engaged segment for reasons that have nothing to do with their interest and everything to do with your calendar.
Once they fall out, they stop receiving campaigns. Once they stop receiving campaigns, they have no chance to open one, and they cannot come back. The segment quietly becomes a trap rather than a filter.
A rough working rule: the window should hold at least six to eight of your sends. Weekly senders land around 60 days. Fortnightly senders want 90 or more. Daily senders can run 30 and it is genuinely engaged rather than an artefact.
What a too-tight segment costs
Revenue that never appears. People who would have bought stop being sent to. There is no line in any report for this, which is what makes it survive for years.
A shrinking base. The segment only refreshes when someone opens. Cut off the sends and the only direction it moves is down, a little every month, which reads as list fatigue and gets treated with more aggressive cleaning.
The wrong conclusion. Revenue drops after the segment is introduced, so the segment gets blamed and the store goes back to sending everything to everyone. That is the worst of both, and it happens often enough to be a pattern.
What a too-loose segment costs
Less than people fear, but it is real.
The complaint rate goes up, because you are mailing people who had drifted off. Sustained, that damages where every email lands, including the ones going to your best customers.
The asymmetry matters: too tight leaks revenue slowly and invisibly, too loose damages deliverability in a way you can measure. So the honest advice is to widen deliberately and watch the complaint rate, rather than to widen and hope.
Opens are a weak signal now, and still the only one you have at volume
Apple Mail Privacy Protection pre-fetches images, so a share of your opens are a machine rather than a person. Anyone telling you to ignore opens entirely has not proposed what to build the segment out of instead.
The practical answer is to keep opens in the definition and to weight clicks above them. A 90-day opener plus a 180-day clicker is a better engaged segment than 60-day openers alone, because a click is a real human and it is worth a longer memory.
What to do this week
Work out your send frequency, set the window to cover six to eight sends, and see how much bigger the segment gets. If it barely moves, the definition was fine. If it grows by a third, you have been sitting on a third of your audience.
Then look at the people who are in the wide version and not the tight one. If they are a group you recognise as customers, the tight version was costing you money.
Should I use 30, 60 or 90 days?
Whichever one covers six to eight of your sends. If you email weekly that is 60. Fortnightly, 90 or more. Nobody can name the right number without knowing how often you send, which is why the number in the article you read was never going to be right for you.
What about people who click but never open?
They exist, because the open is a tracking pixel and the click is not. Any definition built on opens alone will drop them, and they are among the most engaged people you have. Put clicks in the definition with a longer window and the problem goes away.
