A few years ago I logged into an email platform for one of my sites and saw open rates jump overnight. Nothing about the emails had changed. Same subject lines, same send times, same list. I hadn’t done anything differently. The number just went up and stayed up.
That wasn’t good news. It was Apple.
What an open rate actually measures
Open rate tracking is simple and a little crude. Every marketing email you send has a tiny, invisible image embedded in it, usually a single pixel, sometimes literally 1×1 pixels in size. It’s hosted on your email platform’s server. When someone opens the email and their email app loads images, that pixel gets requested from the server, and the server logs it as an “open.”
That’s it. That’s the whole mechanism. It doesn’t know if someone read the email, skimmed it for two seconds, or opened it by accident while cleaning out their inbox. It just knows a request for that image happened. For close to two decades this was a decent, if rough, proxy for “a human looked at this.”
The problem is that the request no longer has to come from a human looking at anything.
The day Apple broke the pixel
In 2021 Apple rolled out Mail Privacy Protection as part of iOS 15, and it later extended to macOS Mail. If a recipient has it turned on, and it’s on by default for a lot of users, Apple’s own servers pre-load every image in every email that arrives, including the tracking pixel, before the person ever opens it. This happens automatically on delivery, routed through Apple’s proxy, regardless of whether the recipient opens the email that day, that week, or ever.
So the “open” gets logged whether or not anyone actually opened anything. It’s not a delay in tracking or a quirk that evens out. It’s a structural change in what the pixel request means. For any list with a meaningful share of Apple Mail users on iPhone, iPad, or Mac, a chunk of your “opens” are Apple’s servers doing a privacy-preserving fetch, not a person reading your email.
This is also why open time became useless around the same period. If Apple’s proxy fetches the pixel near the moment of delivery, “opened at 9:02am” doesn’t tell you when your reader actually looked at the email, if they looked at all.
Why you can’t just subtract it back out
The tempting fix is to say “okay, just discount the Apple opens and look at the rest.” Some platforms do flag MPP-affected opens separately, which helps. But most senders are working from a blended total, and even where the flag exists, you’re only getting Apple’s estimate of who’s protected, not a clean line between real opens and proxy fetches on every single email.
And it’s not a fixed discount you can apply once and forget. The share of your list on Apple Mail changes over time, differs by list, differs by how you acquired subscribers, and keeps shifting as more people update their devices or turn settings on. There’s no stable ratio to divide by. The number isn’t just inflated, it’s inflated by an unknown and moving amount.
It’s not only Apple
Apple is the biggest reason open rate stopped being trustworthy, but it’s not the only one. Corporate email security systems, the kind that scan incoming mail for malware before it reaches an inbox, often pre-fetch images and sometimes links too, as part of scanning the message. That’s another automated “open” that never involved a human.
Pulling in the other direction, plenty of real people still have images blocked by default in their email client, which means a genuine open, someone reading your whole email, can register as nothing at all if they never load images. So the metric is being pushed up by bots and automated scanners on one side, and pushed down by real readers who never trigger the pixel on the other. Whatever number lands in the middle isn’t a clean read on either force.
What that means for subject line testing
If you’ve ever run an A/B test on subject lines and picked a winner based on which one got a higher open rate, that test is now compromised in a way that’s hard to see from the results themselves. A subject line doesn’t influence whether Apple’s proxy pre-fetches an image. It influences whether a human decides to click into the email and, further down, whether they act on it.
If open rate is being padded by automated fetches that happen regardless of the subject line, then differences in open rate between two versions are measuring something closer to noise than to which subject line actually worked. The signal you want, did this subject line get more people to genuinely engage, is still there, but it shows up further down the funnel, not at the open.
What to track instead
Click rate is the more honest signal, because it requires someone to take a deliberate action inside the email. Security scanners occasionally click test links too, so it’s not flawless, but a click is a much higher bar to clear by accident than an image simply loading. If you can see click-to-open rate, meaning clicks as a share of recorded opens rather than of total sends, that’s a useful sanity check on how engaged your genuine readers are, even with a bloated open number underneath it.
Below that, look at what the email was actually for. If it’s a promotional send, conversion rate, meaning purchases or signups that came from that email, tells you what the campaign did. If it’s a relationship-building send, replies are a strong signal, because almost nothing automated replies to a marketing email.
I’d also keep an eye on unsubscribe rate, spam complaint rate, and bounce rate. None of these tell you engagement directly, but they tell you about list health, and a rising complaint rate or bounce rate is worth acting on regardless of what your open rate says.
What open rate is still okay for
I haven’t stopped looking at open rate entirely, I just stopped trusting it for decisions. It’s still a rough early check that an email delivered and rendered, and a number that falls off a cliff can be a useful early warning that something’s landing in spam or a chunk of your list has gone stale, even if the baseline number itself is inflated. Watching the trend within the same list, over a period where your audience’s device mix hasn’t shifted much, can still show you something. What it can’t do anymore is tell you, campaign by campaign, whether people actually opened and read what you sent.
The honest bottom line
Open rate didn’t become a useless number by accident, and it isn’t broken in the sense of being random. It changed what it measures. It used to be a rough stand-in for “a person opened this email.” Now, for a meaningful chunk of most lists, it’s closer to “this email was delivered and an automated process loaded its images.” Those are different things, and treating the second as if it were the first is how you end up congratulating yourself on numbers that moved for reasons that have nothing to do with your writing, your offer, or your list.
None of this means email is a bad channel or that you should stop measuring it. It means the measurement has to move further down the funnel, to clicks, replies, and actual outcomes, where a human still has to do something for the number to move.
If you want more explainers like this on the metrics that actually hold up, take a look at what else we’ve written on Data Research Analysis Collection.
Get new guides and videos first — join the Telegram channel.