The short version
- An open is a loaded pixel, not a reader. Since 2021, Mail Privacy Protection, which the recipient switches on themselves, routes remote content through two relays operated by different entities, loading the images whether the recipient reads the message or not.
- The number is inflated unevenly from one audience to the next: the share of recipients who have switched that protection on varies by segment, so comparisons between segments and between campaigns are distorted, not just the headline level.
- Automations triggered on opens now fire at the wrong people: “non-opener” follow-ups sent to people who did read, flows set off by machines.
- Replacement metrics already exist: the click, the conversion, the value generated per send, and the negative signals, unsubscribes and spam complaints. All of them measure humans.
For twenty years the open rate was email’s headline metric, and it died in 2021 without ceremony. Apple rolled out Mail Privacy Protection: when the recipient switches it on, on iPhone, iPad, Mac or iCloud.com, the remote content of an email, including the tracking pixel, travels through two relays operated by different entities and loads on receipt, not when the message is read. The result is that affected recipients look like they have “opened”, every single time, whether they read the message or binned it.
Four years on, most campaign reviews still open with that number.
What the inflated number actually breaks
| What the open rate is used for | What happens now | The consequence |
|---|---|---|
| Comparing two subject lines in an A/B test | Machine opens spread across both versions and the noise swamps the real gap | “Winners” picked at random, then rolled out everywhere |
| Following up with non-openers | The segment mixes genuine non-readers with readers outside the Apple ecosystem | Follow-ups that feel like pestering to people who already read the message |
| Cleaning the list on inactivity | Genuinely inactive contacts look active, thanks to pre-fetching | A list that is clean on paper and degrading quietly |
| Proving performance to the board | The number has been climbing on its own since 2021 | Campaigns being congratulated while email sales flatline |
The mistake that costs the most Keeping automations triggered on opens: non-opener follow-ups, resends with a different subject line, engagement scoring. They run on data laced with machines, and their misfires are paid for in unsubscribes, the one metric in this channel that has stayed perfectly reliable. Audit your list of triggers before you change anything else.
What still measures a human being
Anything that requires a deliberate action is still true. The replacement hierarchy runs from closest to the reader to closest to the result.
The click is the new floor of engagement: in the vast majority of cases it takes a human hand. It is not perfectly clean either: corporate mail protections, Microsoft Safe Links among them, scan URLs before delivery and can produce clicks nobody made. Its rate is lower than the open rate, which makes it uncomfortable in a board meeting and honest everywhere else.
Conversion and value per send are the real scorecard: what the campaign got people to do, and what it brought in per recipient. This is the email equivalent of what we say about ROAS in advertising: the activity metric flatters, the value metric decides. That does assume the downstream measurement is sound, which brings you back to tracking that lies in silence.
The negative signals, unsubscribes and spam complaints, are the most valuable of all: produced by irritated humans, never by machines, and directly tied to your ability to keep sending at all. A campaign with a “good open rate” and rising unsubscribes is a bad campaign, in that order of priority.
The open rate keeps a residual use: drawing attention to a possible incident. A sudden collapse, measured against your own previous sends, is worth a look. But Google states it plainly: a low open rate does not necessarily indicate a deliverability problem. It is a signal to confirm elsewhere, not proof, and certainly not a steering metric.
How to make the switch, in four moves
- Audit your triggers. List every automation conditioned on an open and rebuild it around a click or a purchase, widening the time window if you need to.
- Change the first column of your reports. Value per send, conversions, clicks, unsubscribes: the open rate drops to the bottom of the page, labelled “technical metric”. Change the table and you change the conversations.
- Redefine what inactive means. A contact with no click and no purchase over a long stretch is inactive, even if they “open” everything. That is the definition that protects deliverability, and it lines up with the basic hygiene we describe in our piece on collecting contacts who have consented: a smaller list that reacts beats a big one that pre-fetches.
- Rerun your subject line tests on clicks. The gaps will be smaller and slower to establish; their advantage is that they will be real.
A ten-minute self-diagnosis: take your last campaign and work out the ratio of clicks to opens. Then compare it with a 2020 campaign if you still have one. The collapse in that ratio, for comparable content, is the share of machines that has crept into your opens. And count your automations triggered on opens: every one of them is currently firing at a mixed target.
The options, and what each is worth
| Option | What you get | Who it is for |
|---|---|---|
| Move reports and triggers onto clicks and value | Lower numbers that are true, and automations that aim straight. An afternoon’s work | Everyone, right now |
| Filter out machine opens | Some tools can tell pre-fetches apart; useful as a supplement, but the filter is partial and the metric stays fragile | Well-equipped teams, as an extra |
| Keep the open rate as an alert signal | Compared against itself, it can draw attention to an incident without proving one: Google points out that a low open rate does not necessarily indicate a deliverability problem | Everyone, at the bottom of the dashboard |
| Carry on as before | Flattering reports, follow-ups aimed at the wrong people, and a list rotting under a coat of good numbers | Nobody |
The tooling
There is no module to install here: any serious sending tool already reports clicks, conversions and unsubscribes. The job is a change of dashboard and of triggers, not a change of tooling. The only technical prerequisite sits downstream: the conversion attributed to email has to be reliable, and that is a matter of how the site measures, not of the sending tool.
The bigger point
The open rate is the textbook case of a metric that outlives its own death because it is comfortable: supplied by default, flattering, familiar. The replacement metrics are lower and slower, so nobody is in a hurry.
It is the same instinct we apply throughout our email marketing expertise: judge the channel on what it gets people to do, not on what it triggers technically. Email remains the most profitable channel in online retail precisely when you steer it on value, because that is where its real strengths, relationship and repeat custom, become visible.
Sources
The verifiable claims in this article link to their primary source, accessed on 11 August 2026. We do not cite a source we have not read.
- Apple, Mail Privacy Protection & Privacy. Open
- Apple, Use Mail Privacy Protection on iPhone (iPhone User Guide). Open
- Apple Newsroom, iOS 15 is available today, 20 September 2021. Open
- Google, Email sender guidelines (Gmail). Open
- Google, Turn images on or off in Gmail. Open
- Microsoft, Complete Safe Links overview for Microsoft Defender for Office 365. Open
FAQ
What share of my opens is machine-generated?
It depends on how many of your recipients have switched Mail Privacy Protection on, a share you cannot see, so it varies by segment and by list: that is exactly why the number is no longer comparable. Some tools will estimate it; the order of magnitude matters less than the conclusion, which does not change.
Is the open rate still useful for deliverability?
Barely. A sudden collapse at one provider is worth a look, but Google notes that a low open rate does not necessarily indicate a deliverability problem: the signal has to be confirmed elsewhere. Mailbox providers themselves judge your sends on complaints, deletions without reading and genuine engagement, not on your pixel.
Should I throw out all my subject line tests?
The ones judged on opens since 2021, yes, unless the gaps were enormous. Run them again on clicks, with bigger samples and longer windows. A subject line is judged on what it gets clicked and bought, which tests the whole message, not just its first line.
What should my flows trigger on, if not opens?
On actions: a click, a site visit, a purchase, and the absence of those over wider windows. A win-back flow conditioned on “no click and no purchase in six months” targets genuinely inactive contacts; conditioned on “no open in three months”, it targets mixed data.
Can Dotsland help with this?
Yes. Auditing your triggers, moving reports onto value per send, redefining inactivity and making attributed conversion trustworthy. It is one of the projects covered by our email marketing expertise. Let’s talk, or start by counting your automations triggered on opens.