Email Open Rate Is Not Measuring What You Think

Email open rate is not a reliable signal of program health, and for most inboxes it has not been one for years. Apple Mail, including Mail Privacy Protection, now accounts for 62.26 percent of tracked email opens, according to Litmus's most recent client market share data.¹ Litmus does not equivocate about what that means: opens routed through Mail Privacy Protection are not considered reliable opens. At that level of contamination, the number no longer measures what it claims to measure.

Mail Privacy Protection does more than hide who opened a message. It removes the concept of opening from the signal entirely. Apple's own documentation describes a two-relay proxy: the first relay sees a device's IP address and nothing else, the second sees the remote content Mail downloaded and nothing else, and no single party holds both.² The content, including a sender's open-tracking pixel, downloads in the background as soon as a message arrives at the device, regardless of whether a person ever looks at it. The pixel fires, but nobody is there.

That is different from undercounting. Undercounting is a bias a program can correct for. A fabricated open event that fires on delivery rather than engagement is noise indistinguishable from signal at the individual message level. Segmentation built on open behavior, send-time optimization tuned to when subscribers historically opened, and A/B tests that pick a winning subject line by opens are all running on a number a server relay generated, not a subscriber.

The biggest objection is that click-through rate carries its own contamination. Corporate security gateways and some mail clients pre-fetch links to scan them for malware before a subscriber ever clicks, which inflates click counts the same way Mail Privacy Protection inflates opens. That objection is correct, and it does not rescue open rate. A metric with partial, variable noise a program can model against still beats one that fires on delivery for well over half a list by design. The question was never which metric is perfectly clean. It is which one still correlates with a human being on the other end.

Think about conversion instead. Where conversion is too sparse to be statistically useful, measure clicks against delivered messages rather than against opens, so the denominator is one Mail Privacy Protection never touched. Deliverability signals that never depended on opens in the first place (bounce rate, complaint rate, and inbox placement) say more about list health than an open count ever did. Email open rate can stay on the dashboard, as one noisy input among several and weighted accordingly, as long as nobody treats it as ground truth.

Most marketing teams have not rebuilt their dashboards around that shift yet, which means most program health decisions being made this quarter are still keyed to a number Apple manufactures on a server before anyone reads the message.