Why email open rates lie, and what to measure instead

An email “open” is recorded when a tiny tracking image is requested, not when a person reads the message. Apple Mail Privacy Protection loads that image automatically for recipients who switch it on, so Apple Mail users can show as opens without reading anything. Treat open rate as a rough signal at best. For cold email, decide with positive replies, and watch bounces, spam complaints and unsubscribes as the health checks.

This article explains what an open is, what Apple’s own documentation says about Mail Privacy Protection (MPP), why the open rate cannot be repaired by filtering, and which numbers to use.

How an open is measured

The sender adds a one-pixel image to the HTML message. The image’s address contains a token that identifies the message. When the recipient’s mail app or a mail server downloads the image, the sender’s server logs a request and calls it an open.

That is the whole mechanism, and it has three weak points:

• No image request, no open. A recipient who reads the message with images blocked, or reads the plain-text version, is not counted.
• An image request is not a person. Anything that downloads the image counts, whether or not anyone reads the message.
• The first request can be anything. The time, place and device behind it are the requester’s, not necessarily the reader’s.

What Apple says Mail Privacy Protection does

Three Apple primary pages, read on 2026-10-05, describe it in Apple’s words.

1. Apple Support (Mail on iPhone): Mail Privacy Protection “hides your IP address so senders can’t link it to your other online activity or determine your exact location”. It also “prevents senders from seeing if you’ve opened the email message they sent you”. A note says it does not extend to links or attachments. It is switched on in Settings, under Mail and Privacy Protection, as Protect Mail Activity.
2. Apple’s iOS 15 announcement (June 2021): it “stops senders from learning whether an email has been opened”.
3. Apple’s WWDC21 session on privacy: Apple tells developers that if they use remote images to measure campaigns, there are changes to be aware of. Mail content may be loaded automatically after delivery, so the time of viewing is no longer correct. The content is loaded without revealing IP addresses and without detailed headers, so the location and type of device are not revealed. And “you’ll see your emails as being opened, regardless of if the user read it or not.”

Read the last sentence twice. It comes from Apple, and it describes the mechanism that inflates the number.

Two limits on what we can say. Apple’s pages describe the feature as something people turn on, and we did not find a primary figure for how many recipients have it enabled or how many Apple Mail users there are in your list, so we do not give a share. And Apple’s note that links are not covered means clicks are not hidden by MPP, but we do not claim clicks are clean, because other tools can request links too, and we have not verified how.

Google’s sender guidelines say the same from the mailbox provider’s side: Google does not track open rates, it cannot verify the accuracy of open rates reported by third parties, and low open rates are not necessarily an accurate indicator of deliverability or spam classification issues.

What this does to your numbers

The effect depends on the share of your list that reads mail in Apple Mail with the feature on. That share is unknown to you, differs by audience, and changes over time, which is why you cannot correct for it with a formula.

A made-up example, to show the logic and not to represent a benchmark:

Assumed: 1,000 messages delivered, 300 recipients in Apple Mail with MPP on
Image requested for all 300 automatically:        300 "opens"
Other recipients who really opened and loaded:    150 opens
Reported open rate:                               (300 + 150) / 1,000 = 45%
Real human opens, if you could count them:        unknown: it is not 150 and not 450

The 300 could include people who read, and people who did not. You cannot separate them from the sender side. Change the assumed Apple share from 30% to 50% next quarter and the open rate moves without any change in the message, subject or audience quality. A test that compares two subject lines on open rate is measuring, in part, which version happened to go to more Apple Mail users.

Three consequences for cold email:

1. Subject-line tests on opens are not valid. The winner may be the version with more automated opens.
2. Open-based triggers misfire. “Send a follow-up only to people who opened” treats automated opens as interest.
3. Open rate cannot diagnose deliverability. A message in spam can still be recorded as an open if the image is fetched, and a message in the inbox can show no open if images are blocked.

Why filtering does not fix it

Tools can try to discard opens that look automated, for example an open within seconds of delivery or one from a known prefetch client. A heuristic like that helps, and it remains a heuristic. It cannot recognize an automated request that looks like a human one, and it can discard a real, fast human open. Use it to clean the chart, not to make decisions.

What to measure instead

Pick numbers that need a person to act, or that the mailbox providers themselves watch.

Keep three habits:

• Divide by delivered, not sent, so that bounces do not hide inside the denominator.
• Classify replies. An automatic out-of-office and “stop emailing me” are replies and are not interest. Count them separately.
• Read ratios with their counts. “3% positive” of 40 delivered is one reply. See how much data you need.

For deliverability without opens, use the signals mailbox providers publish: the spam rate and compliance status in Google Postmaster Tools, delivery errors, and your own bounce codes. A seed test before a launch shows where a message lands in a sample of inboxes and is a snapshot, not a measure of your audience. The deliverability audit checklist puts these together.

Should you track opens at all?

You can, as long as you use the number for what it is. Our default is conservative:

• Off on the first message. A tracking image adds a remote URL and a domain to the first impression, which is the message where it costs the most.
• If on, on a domain you control with its own SPF and DKIM, so the image does not borrow someone else’s reputation.
• Never in decisions. No winner selection, no follow-up logic, no mailbox health verdict from opens.

How Dooxout handles this

Dooxout’s metrics lead with delivered, replied, bounced, unsubscribed and complaints, and A/B results are decided on positive replies, not on opens. Open tracking is a per-campaign switch and is off on the first step by default. Where tracking is on, the dashboard labels the open figure as approximate because of Apple Mail Privacy Protection, flags opens it detects as automated with a simple heuristic (an open within seconds of delivery or a known prefetch user agent), and leaves those out of the unique-open headline while keeping them in the raw data.

Replies are classified (reply, out-of-office, bounce, unsubscribe), so an out-of-office message does not count as interest. The platform shows the numbers and warns; what you decide to do with them is up to you. See email deliverability and integrations for how reports and events reach your CRM.