Apple Mail Privacy Protection Arrived in 2021 and the Open Rate Has Meant Nothing Since

What MPP Actually Does

Apple's Mail Privacy Protection launched with iOS 15, iPadOS 15, and macOS Monterey in September 2021. The feature's mechanics are straightforward and their consequences are severe. When a user opts into MPP — and Apple's own opt-in prompt was phrased in a way that nudged strongly toward acceptance — Apple's servers pre-fetch the content of every incoming email, including any tracking pixels embedded by the sender's platform, before the user ever opens the message. The IP address recorded is Apple's proxy, not the recipient's. The timestamp is Apple's retrieval moment, not the reader's.

Tracking pixels have been the measurement foundation of email marketing since the late 1990s: a one-by-one-pixel invisible image, the load of which signals an open. MPP severs that signal entirely. A pixel loaded by Apple's infrastructure at 3 a.m. looks, from the sender's analytics dashboard, identical to a deliberate human open at 9 a.m. There is no way to distinguish them in aggregate data.

The Measured Collapse of a Metric

The numbers that emerged from the months following September 2021 were unambiguous. Litmus, which tracks email client market share through its panel data, observed Apple Mail's share of the email client market at roughly 50 percent or higher among many publisher audiences — meaning MPP's reach was immediate and enormous. Mailchimp published analysis showing sharp, discontinuous jumps in reported open rates for lists with large proportions of Apple Mail users, with some senders seeing measured opens climb 20 to 30 percentage points almost overnight. The jump was not a readership surge; it was inflation, the term the industry settled on to describe the statistical artefact.

beehiiv, the newsletter infrastructure company founded by former Morning Brew employees, has written extensively about how its publisher data reflects the same pattern: open rate figures for Apple Mail users can no longer be treated as behavioural signals. Sending-time optimisation algorithms — tools that attempt to deliver messages when a reader is most likely to open them — became unreliable for any portion of a list using MPP-affected clients, because the timestamps feeding those algorithms are now Apple's prefetch schedule, not human behaviour.

The metric that newsletter operators had used as their primary performance signal since the format began — more reliable, supposedly, than the click-through rate, and certainly more abundant — became noise. Not partially degraded noise. Fundamentally untrustworthy noise, at scale, from the dominant email client.

What Survives

Click-through rate survives. A click requires a human action that Apple's pre-fetching cannot simulate. Reply rate survives. Conversion data, where a reader arrives at a destination and completes a tracked action, survives. Some platforms shifted to reporting "click-to-open rate" calculated only on Gmail and other non-MPP clients, though the sample-selection problem this introduces is obvious: Gmail readers are not a representative slice of any general audience.

The episode carries a structural lesson that extends beyond email measurement. When a single platform controls the delivery layer for a large fraction of an audience, that platform can — deliberately or as a side effect of a privacy-motivated feature — reshape how the entire publishing ecosystem understands its own reach. Apple's stated rationale for MPP is user privacy, which is a legitimate interest. The effect on publishers is collateral damage. Whether that framing is accurate or convenient is a question worth holding. What is not in dispute is the outcome: the open rate, as a signal, broke in September 2021 and has not been fixed, because it cannot be fixed within the current architecture. Platforms that built engagement dashboards, content strategies, and even pricing tiers around open-rate benchmarks have had to quietly acknowledge that the primary number on their reports is now, at best, a rough ceiling and not a floor — and at worst, an artefact of Apple's server schedule.