Apple MPP Broke Your Open Rate. Four Metrics That Still Tell the Truth.

Apple Mail Privacy Protection pre-loads every tracking pixel, counting an email as opened whether or not a human ever saw it. One study found total open rates jumped from 22.6% to 40.5% after MPP rolled out, nearly doubling without a single extra person reading anything. With Apple Mail sitting near half of all recorded opens, your open rate is not a metric anymore. It is noise. The signals that still track real humans are clicks, conversions, replies, and list health.

The number that broke, and why it took the others with it

Fast mechanics, because the fix depends on understanding them. Mail Privacy Protection routes messages through Apple's proxy servers, which fetch every image and tracking pixel the moment an email lands, before anyone taps it. That pixel is exactly what your ESP uses to register an open. So Apple fires it for everyone on Apple Mail, engaged or asleep or dead-list, and your reporting can't tell the difference.

The distortion isn't subtle. Omeda compared the two months before MPP to the period right after and watched total open rate climb from 22.6% to 40.5%, with unique open rate going from 15.2% to 29.0%. Both nearly doubled while nothing about the actual audience changed. Paubox pegs the inflation at 15 to 20 percentage points and estimates Apple Mail now accounts for close to half of every open you record. When half your data is machine-generated, the average means nothing, and neither does any trend line you draw through it.

Here's the part most teams miss, and it's the reason a simple "track clicks instead" answer isn't enough. Open rate didn't break alone. It dragged click-to-open rate down with it. CTOR is clicks divided by opens, so if the denominator is stuffed with robot opens, the ratio collapses for reasons that have nothing to do with your subject line or your creative. If you're still putting CTOR in a client deck in 2026, you're grading yourself against a ruler that changed length halfway through the year. Personally, I'd retire it quietly this quarter and not mention it again.

Half your open rate is a proxy server in Cupertino loading a pixel while your subscriber sleeps.

Clicks are the floor you build everything else on

The good news is boring, which is usually how good news arrives in deliverability. Clicks survived. MPP pre-loads images, but it doesn't click links for people, so total clicks and unique click rate are still measuring a deliberate human action. Beehiiv's own guidance to publishers is blunt about this: opens are compromised, clicks are not, so compare clicks across sends of similar size and you'll actually learn which emails landed.

Benchmarks, so you have something to aim at. A click rate of 2-5% is healthy for most newsletters. Media and creator-led sends can push 3-6%. B2B tends to sit lower, in the 1.5-2% range, because the audiences are smaller and the asks are heavier. These aren't laws of physics, and your list's history matters more than any cross-industry median, but if you're a consumer newsletter running under 1.5% clicks on broadcasts, something upstream is wrong. It's usually either the offer or the list, and those are two different problems (more on the list one below).

One caution on clicks, because it's the mistake I see most: bot clicks are a thing too. Corporate security scanners and link-protection tools fire clicks the same way MPP fires opens, especially in B2B. If you see a burst of clicks within seconds of send, from the same user agent, across every link including the unsubscribe, that's a scanner, not a reader. Most decent ESPs filter these now, but check before you celebrate a 9% click rate on a cold B2B send. It's probably a firewall admiring your work.

Stop segmenting on opens. Your engaged list is defined by actions now

This is the change that actually moves revenue, and most teams still haven't made it. For a decade, "engaged" meant "opened in the last 30 or 90 days." That definition is now garbage, because a subscriber who never once looked at your email still counts as a 40-times-in-90-days opener thanks to MPP. Re-engagement campaigns built on open history are re-engaging people who were never disengaged, and worse, they're leaving genuinely dead addresses on the list.

Rebuild the segment on things a human has to physically do. A practical starting definition, and one data teams are converging on post-MPP: engaged means clicked at least once in the last 120 days, or completed a conversion in the last 180 days. Everyone else drops into a sunset track. You don't delete them immediately. You send a short "still want this?" sequence, and if they don't click, you stop mailing them.

Why bother, beyond tidiness? Deliverability. Gmail and the other mailbox providers watch how your recipients behave, and mailing a large block of people who never engage drags your sender reputation down for everyone, including the subscribers who do want your email. This connects directly to how Gmail decides Promotions-tab placement, and it compounds with your authentication setup. You can have perfect SPF, DKIM, and DMARC and still land in spam if you keep hammering a list of ghosts. And remember that the list is decaying underneath you the whole time. Roughly a fifth of a typical list goes stale every year, so an action-based sunset policy isn't hygiene theater, it's the thing keeping your reputation alive.

The action: pull your list today and re-cut your engaged segment on clicks and conversions only. Anyone who's clicked in 120 days or bought in 180 stays. Everyone else gets one sunset sequence, then goes quiet. Expect your "engaged" count to shrink, maybe a lot. That's not you losing subscribers. It's you finally seeing how many you actually had.

The two metrics almost nobody puts on a dashboard

Clicks and conversions are the obvious survivors. The interesting ones are further down, and they come out of Validity's 2026 benchmark work, which MarTech summarized into three relationship-health metrics. Two of them are worth stealing.

First, reply rate. Almost no marketing team tracks it, because most marketing email is sent from a no-reply address, which tells you something about how the industry thinks about the reader. A reply is the highest-intent signal a subscriber can send you short of buying. Validity's read is that even a 1% reply rate represents real audience investment, and mailbox providers treat replies as a strong positive trust signal that helps inbox placement over time. So the move is small and slightly counterintuitive: send from a real, monitored address, and every so often ask a genuine question that invites a response. In your welcome sequence especially, one "just hit reply and tell me what you're working on" email does double duty. It builds trust with a human and teaches Gmail you're a real correspondent, not a broadcast tower.

Second, the disaffection index. Validity's version combines unsubscribes, complaints, and bounces into one number that answers a question open rate never could: how fast are you burning through your audience? I like this one because it's honest in a way vanity metrics aren't. You can juice an open rate. You can't fake a low complaint rate while annoying people. Track unsubscribes plus spam complaints plus bounces as a single percentage of delivered, watch the trend, and treat any sustained climb as the smoke alarm it is. A rising disaffection index three sends before your open rate would've told you anything is the early warning that used to be invisible.

Rebuild the dashboard: four numbers and one afternoon

Here's what I'd actually put in front of a team, and it fits on one screen. Delete open rate from the top-line view. Not because opens are worthless forever, but because leaving an inflated number at the top of the dashboard guarantees someone optimizes for it. Keep it in a footnote if you must, labeled "MPP-inflated, do not use for decisions."

The four that go up top: click rate (your engagement floor), conversion rate (the only metric your CFO respects), reply rate (trust and deliverability in one), and the disaffection index (your burn rate). If you want a fifth, add list growth rate net of the sunset track, so you're watching real, engaged size rather than a bloated total. Rebuilding this in most ESPs is genuinely an afternoon of work: re-cut the engaged segment, unhide the reply-from address, build one saved report. You'll know more about your list by Friday than you did all year.

Two questions I get asked about this

Should I still A/B test subject lines if I can't trust opens? Yes, but change what you measure. Test the subject line against downstream clicks or conversions, not opens. It's a slower signal and you need a bigger sample, so run subject tests on your engaged segment where the click volume is real. Testing subject lines on open rate now is testing which version Apple's proxy prefers, which is to say nobody's preference.

Is open rate completely useless, or just mostly? Mostly, with one exception. For the slice of your list on non-Apple, non-proxied clients, opens are still roughly real, so a segment-level open rate filtered to those clients can be a directional signal. It's fiddly, most teams won't bother, and I don't think you need it if you're tracking clicks properly. But if you're the kind of person who likes a clean dataset, that's where the last honest open rate lives.

The teams still reporting a 42% open rate to their board are going to look great right up until someone asks what it's driving. My honest guess is that within a year, "open rate" quietly disappears from most serious email dashboards the way "hits" disappeared from web analytics twenty years ago. Not because anyone banned it. Because it stopped answering a question worth asking, and the people who moved to clicks, replies, and revenue simply stopped needing it.

Notice Me Senpai Editorial