Apple Proxy

Deliverability Case Study: "Apple Proxy"

Hendrix's "Foxy Lady" is a leering, fuzz-heavy 1967 hard-rock come-on, and the parody keeps the swagger but turns the object of desire into a number: the open rate. The sender is infatuated with a metric that looks devoted and is mostly a machine. The "Apple Proxy" is Apple Mail Privacy Protection (MPP), and the song follows the sender from seduction to disillusion to a decision to chase something real.

Here is the technical breakdown of the deliverability concepts detailed in the song:

Verse 1: The Open Faker

"You know you're a cold little open faker / ... / And you know you're a fake little stat-maker"

• The Deliverability Context: Open tracking works by embedding a tiny remote image (the tracking pixel) in each email. When the image is fetched from your server, your Email Service Provider (ESP) records an "open." MPP, launched with iOS 15 and macOS Monterey in September 2021, breaks that link. When a user enables it in the Apple Mail app (and most do, because Apple offers it at setup), Apple's proxy servers download all remote content in the background, usually soon after delivery, whether or not a human ever looks at the message. Your pixel fires and your ESP logs an open. "Cold" is accurate: no one is reading. "Faker" is fair too, because the open is real at the HTTP level and means nothing at the human level.
• The Scope: MPP applies to the Apple Mail app, whatever the account is. A Gmail or Outlook address read in Apple Mail on an iPhone is proxied, but the same address read in the Gmail app is not. Your inflated audience is "everyone who reads in Apple Mail," which for many consumer lists is a large share.

Chorus 1: "I Want to Know Who Read"

"I want to know who read, uh huh, yeah / These opens don't mean a thing, no, ha!"

• The Deliverability Context: That is the one thing MPP is designed to deny you. Because the proxy fetches the pixel for every message delivered to a protected inbox, an MPP open tells you only that the email reached an Apple Mail client. It does not tell you it was read, when it was read, or whether it was read at all. The proxy also masks the recipient's IP address, so open-based location and device data degrade to a coarse region.
• The Correction Senders Need: "Don't mean a thing" goes slightly too far. An Apple proxy open does confirm delivery to that client, which is weak evidence that the message was not rejected. It says nothing about engagement, though, and the chorus is right that it should never be read as interest.

Verse 2: Skewing the Scene

"Now I see you, heh, are skewing the scene / ... / You make me want to get up and scream"

• The Deliverability Context: The damage does not stay inside one dashboard number. Everything downstream that keys on opens is skewed:

- Inactivity and sunset rules: an Apple Mail subscriber who hasn't read anything in a year still "opens" every send, so open-based suppression never removes them. Your "engaged" segment fills with people who are not engaged. - Subject-line A/B tests: picking a winner on open rate is close to random when a large share of opens in both variants are automatic. - Open-triggered automations: "opened but didn't click" follow-ups and send-time optimization fire on proxy behaviour, not human behaviour.

• Why It Matters for Placement: Mailbox providers do not see your ESP's open pixel data. Gmail and the others filter on their own internal signals. MPP inflates your view of engagement, not your reputation. The risk is indirect: you keep mailing people who don't want the mail, because your data says they do, and then the complaints and ignored messages that the providers can see will catch up with you.

Chorus 2 & 3: "Show What's Real" and "Track You Down"

"I've made up my mind / I'm tired of wasting all my precious time / You've got to show what's real, what's real" "Yeah, I'm gonna track you down, haha / These opens don't prove a thing, no"

• The Resolution: "Show what's real" is the correct pivot. Measure what a proxy cannot fake: clicks, click-to-open rate (CTOR) only where opens are non-proxied, replies, conversions, and on-site activity tied to the email.
• On "Tracking You Down": Most ESPs now try to flag proxy opens, using Apple's proxy IP ranges and fetch patterns, and report them separately as "machine opens." Use that flag to exclude MPP opens from engagement logic. Don't treat it as a way to recover who actually read. Trying to fingerprint past a privacy feature the user switched on is not a deliverability tactic. It is a consent problem, and in jurisdictions where open tracking already requires consent, it compounds the problem.

The outro's "Oh, you mess my stats" is the honest verdict: MPP didn't break email, it broke one metric. The sender who stops loving the open rate and starts measuring action gets a truer picture of the list, and a list that is actually healthier.