That open didn't happen.Half the opens in your report might be ghosts. Nobody read the email. A server just fetched a picture. Here's how caching quietly turns your open rate into fiction.
⬡ myth-fact, MYTH "an open means someone read it" vs FACT "an open means a pixel loaded, and a pixel can load itself"Open tracking is one tiny trick. Your email carries an invisible one-pixel image. When someone opens the message, their app loads that image from your server, and you log an open. Simple in theory. In practice, caching snaps the wire.
⬡ dissect, label the invisible tracking pixel inside a mock email bodyApple opens it before the human does.The loudest example is Apple Mail Privacy Protection. Apple's servers pre-fetch every image in the message, your tracking pixel included, often before your subscriber has even tapped the email. So every Apple Mail user shows up as an opener, whether they opened it or not. Captain Kraken never touched your newsletter, but your report swears he did.
⬡ journey-flow, envelope routes through an Apple proxy node that fires the pixel, then arrives unopened at the inboxGmail caches too, through its own image proxy, but it behaves the other way around. It caches after the first open, not before. So it doesn't inflate your counts like Apple does. The catch: a second look at the same email from the same person might not register, because Gmail serves the stored copy and never pings you again.
⬡ split-compare, LEFT "Apple: caches BEFORE open, inflates" vs RIGHT "Gmail: caches AFTER first open, hides repeats"On a list heavy with Apple Mail, this isn't a rounding error. Open rates can read twenty to forty points higher than real human engagement. An honest fifty percent might really be thirty. You simply can't treat the open count as a headcount of readers anymore.
⬡ talking-stat, big "+20 to 40 pts", pill: "inflation on Apple-heavy lists"If this is untangling your metrics, subscribe, we go deep on the whole trust-your-data playbook.
⬡ title-card, on-screen "Subscribe · trust your data"So what can you trust? Clicks. A click needs an actual finger on an actual link, so caching doesn't fake it the way it fakes opens. Bots are a separate headache we cover elsewhere, but for real human interest, weight clicks over opens every time.
⬡ gauge-meter, needle: "opens" reads unreliable, "clicks" reads reliableUse clicks as your primary signal for segmenting and for re-engagement thresholds. Opens are still useful for direction, a rough sense of interest, but never suppress someone or re-engage them on opens alone.
⬡ checklist, "Clicks = primary signal · Opens = directional only · Don't suppress on opens alone"Next up: the mechanics of Apple's Mail Privacy Protection, and exactly how it biases your engagement data. Subscribe and take that one next.
⬡ end-card, Subscribe + Next: "How MPP caching biases engagement data" (004.010.004)How does email client caching affect metric accuracy? Your open rate depends on a tiny tracking pixel loading when someone reads your email. Caching breaks that. Apple Mail Privacy Protection pre-fetches every image through Apple's proxy before the subscriber even opens the message, so every Apple Mail user counts as an opener. Gmail caches the other way, after the first open, which quietly hides repeat opens. On Apple-heavy lists this can push reported open rates twenty to forty points above real engagement. Clicks stay far more reliable because a click needs a real human. This video shows why, and how to lean on click data for segmentation and re-engagement instead of trusting opens alone.
0:00 The open that never happened
0:20 How the tracking pixel works
0:45 Apple's proxy fires it early
1:20 Why Gmail behaves differently
1:50 How big the distortion gets
2:20 Trust clicks, not opens
Next: How MPP caching biases engagement data → [link 004.010.004]
Full written guide → reviewmyemails.com/emailalmanac
#email #deliverability #emailmetrics