Most lists run on one split: people who open and people who don’t. Openers get the campaigns, everyone else gets the win-back, and the strategy conversation ends there. If your email segmentation still starts and ends with openers vs non-openers, you’re sorting subscribers by a number their mail app invented.
That number broke in September 2021, when Apple shipped Mail Privacy Protection. Apple Mail now loads tracking pixels on Apple’s servers whether the subscriber reads the email or not, and every one of those loads registers as an open. Corporate security filters inflate the count further when they scan inbound mail. The metric still sits in every dashboard, but it mixes real readers with machines, and your platform can’t tell you which is which.
The fix is to rebuild segments on data the platform can verify. The rest of this post is that rebuild: which data to trust, and five segments worth building this week.
Why opens stopped being a signal
Mail Privacy Protection ships with Apple Mail, one of the most used email clients in the world. When it’s on, Apple pre-fetches the images in the email, including the invisible pixel your platform uses to record an open. The subscriber may never have seen your subject line. The open counts anyway. Adoption spread fast because it’s a one-tap prompt, so this isn’t an edge case on consumer lists.
Aggregate open rates still carry some directional value, like whether a subject-line test moved anything across a large send. As a per-subscriber segmentation key, opens fail exactly where you need them: you can’t look at one contact’s history and know whether a person or a server produced it.
The cost shows up on both sides of the split. Define “engaged” as “opened in the last 90 days” and your campaign audience is padded with phantom readers who never see the email. Meanwhile, real customers whose mail clients under-report opens get shunted into win-back sequences they never needed. Both mistakes leak revenue quietly.
Segment on what people do, not what their mail app reports
Every subscriber action sits somewhere on a reliability scale. Purchases are at the top: money moved, and your store recorded it. Site visits that start from an email are nearly as reliable. Clicks come next; corporate link scanners fire some phantom clicks, but far fewer than pixel pre-fetching fires phantom opens, and most platforms now filter the obvious bot clicks out. Opens sit at the bottom of the scale.
So the working rule for email segmentation in 2026: build segments from clicks, site activity and purchase data first, layer in what subscribers have told you directly, and treat opens as tie-break information at most.
None of this works without the plumbing. Three inputs power the whole model: your platform’s onsite tracking snippet for site visits, tagged links so clicks attribute cleanly, and a connected store feed for orders.
Ten minutes checking those three saves a quarter of arguing with bad data.
The four data layers that predict revenue
Four layers cover almost everything a campaign needs to know about a subscriber.
- Engagement recency: when did this person last click an email or visit the site? A simple 30, 60 and 90-day banding on those two actions replaces the openers list outright.
- Purchase history: never bought, bought once, bought twice or more. The jump from first to second purchase is where DTC margin lives, so this layer decides more revenue than any other.
- Lifecycle stage: your automated flows already own the lifecycle: welcome, post-purchase, win-back. Segments exist so campaigns don’t collide with those flows, like excluding anyone still inside a welcome sequence from a promo send.
- Declared data: what subscribers told you at signup: category interest, buying timeline, how often they want to hear from you. It’s the only layer people hand you directly, and most brands never ask for it.
The layers multiply rather than stack. “Clicked in 90 days” crossed with “bought once” names the exact audience for your next cross-sell campaign, which no single metric could find on its own.
Five segments to build this week
You don’t need forty segments. Five do the early work:
- Engaged 90: clicked an email or visited the site in the last 90 days. This becomes your default campaign audience, replacing “all subscribers”.
- VIPs: two or more purchases, or lifetime spend in your top band. They’ve proven they don’t need a coupon, so send them early access and new products before anyone sees a discount.
- One-time buyers in the repeat window: bought once, 30 to 90 days ago. The highest-leverage cross-sell audience on your list.
- Warm non-buyers: clicking regularly, never purchased. Some objection is holding this group back, which makes it the one segment where a first-order offer earns its cost.
- Sunset: no click or site visit in 180 days, after at least one re-engagement attempt. Suppress them from regular sends.
Treat the thresholds as starting points. A coffee brand should tighten every window; a mattress brand can loosen them. Set them from your own purchase cycle, then leave them alone long enough to read the results.
One build note: make the segments mutually exclusive at send time. A VIP who clicked yesterday also qualifies for Engaged 90, and without exclusions they get both versions of the same campaign. Most platforms handle this with exclusion rules stacked in priority order: VIPs first, then one-time buyers, then the rest of Engaged 90.
Why suppressing subscribers makes you more money
The sunset segment is the one brands resist, because list size feels like an asset. Deliverability works the other way. Gmail and Outlook score your sending reputation largely on how recipients interact with your mail. Send to everyone and the unengaged majority drags that score down, which pushes your campaigns toward the spam folder for the whole list, buyers included.
Suppress the dead weight and the engaged share of every send rises, inbox placement improves, and the same campaigns reach more of the people who actually spend. The list looks smaller in the dashboard and behaves bigger at the register. A subscriber who hasn’t clicked in six months and ignored a re-engagement email costs you inbox placement with the customers who do buy.
Give people one honest exit before you suppress them. A short re-engagement email asking whether they still want to hear from you beats a month of pretending; whoever clicks re-enters Engaged 90, and silence is your answer for the rest. You don’t need to delete anyone. Suppression keeps the record and the consent while taking dead addresses out of the daily reputation equation.
Judge every segment by revenue per recipient
Open rate can’t referee this system, so use revenue per recipient: campaign revenue divided by the number of people who received it. Say a blast to 20,000 subscribers books $4,000. That’s 20 cents per recipient. The same campaign to 6,000 engaged subscribers books $3,000, which is 50 cents. Total revenue fell and the programme still got stronger, because the second send made more from every address it touched and spent none of your sender reputation on dead inboxes.
Run that comparison over four to six campaigns before you judge a segment; single sends are noisy. Then the harder test: a segment earns its place when it changes what you send. A different offer, a different frequency. If two segments receive identical campaigns every time, merge them and spend the effort where behaviour actually differs.
Want email that behaves like a revenue channel?
Segmentation is one part of the retention system we build and operate, alongside flows, campaign strategy and deliverability. See how we run email marketing as a system, or book a call and we’ll map your list against the five segments above.