Which email metrics to trust now that opens are unreliable
Why Apple Mail Privacy Protection, image proxies and link scanners distort opens and clicks, and which numbers to report on instead.
Opens no longer tell you who read your email. Apple Mail loads images for many readers whether they read or not, other providers fetch images through proxies, and many people block images. Trust the metrics that need a person to act or that a mailbox provider reports back: clicks (after filtering out scanners), conversions and revenue, replies, unsubscribes, complaints, bounces, and results split by recipient domain. Use opens only to compare similar sends with each other.
How an open is counted
An email platform counts an open when the reader’s mail app downloads a tiny image, a tracking pixel, that is unique to the recipient. No image download, no open. An image download by something other than the reader, still an open.
Three things now break that link between the download and a person reading.
Apple Mail Privacy Protection
Apple introduced Mail Privacy Protection in iOS 15 in 2021 (Apple, WWDC21). When a user turns it on, Apple Mail loads remote content privately, hides the user’s IP address, and, in Apple’s words, “prevents senders from seeing if you’ve opened the email message they sent you” (Apple Support).
In practice, a protected Apple Mail user shows as opened whether they read the email or not. The open time and location are not theirs either. Apple’s note also says the protection “does not extend to links”. Clicks still come from the reader.
Image proxies
Since 2013, Gmail has served images through Google’s own proxy servers instead of loading them straight from the sender’s server (Gmail blog). The sender still sees an open when the image is fetched, but the IP address and location belong to Google, not the reader. Device and location reports built on opens are unreliable for Gmail readers.
Images turned off
Some readers and many company mail setups block images by default. A person can read every word and never count as an open. So opens overcount Apple Mail users and undercount people who block images.
Clicks are better, but not clean
A click needs a link request, which Apple’s protection does not fake. Clicks are the best engagement signal you have. They still carry noise from security software.
Some company mail gateways and security tools follow links in incoming email to check them for malware before the reader sees the message. Those visits count as clicks. The unsubscribe standard RFC 8058 exists partly for this reason: it notes that “anti-spam software often fetches all resources in mail header fields automatically”, and that a sender cannot tell an automatic request from a person (RFC 8058). That is why one-click unsubscribe uses a POST request, not a plain link visit.
Signs that clicks came from a scanner:
- Every link in the email clicked by one recipient.
- Clicks within seconds of delivery.
- Many clicks from addresses at one company domain at the same moment.
- Clicks with no visit, signup or purchase on your site afterwards.
You don’t need to filter every one. Judge a campaign by clicks on the main link and by what happened after the click, and check results by recipient domain when one company’s numbers look strange.
The metrics to use
| Metric | What it tells you | Trust |
|---|---|---|
| Conversions and revenue | The email caused the action it was for | High, when tracking is set up |
| Clicks on the main link | A person, or a scanner, wanted more | Good, after a sanity check |
| Replies | A person read it and cared enough to write | High, but only for emails that invite replies |
| Unsubscribes | The email or the frequency stopped being wanted | High |
| Spam complaints | The recipient no longer wants it and said so to the provider | High |
| Bounces | Addresses that don’t exist, or a provider refusing your mail | High |
| Results by recipient domain | Whether one provider treats you differently | High |
| Opens | A rough trend between similar sends | Low |
Conversions and revenue
Decide what the email is for before you send it: a signup, an upgrade, a purchase, a feature used. Then count that action. EmailMarketer.io shows conversions and revenue in the campaign report when you track them, so the result sits next to clicks and unsubscribes.
Unsubscribes and complaints
These are the most honest signals in email, because a person has to act. Gmail asks senders to keep the spam rate shown in Postmaster Tools below 0.10% and never reach 0.30%. Yahoo’s limit is also 0.3%. A rise in unsubscribes after a change in frequency or content usually comes before a rise in complaints. Our post on spam complaint rates covers the limits in detail.
EmailMarketer.io processes complaints from feedback loop reports by unsubscribing and suppressing the address, and shows bounce and complaint rates on each campaign, with a warning when they run high.
Bounces
Hard bounces mean the address doesn’t exist. EmailMarketer.io suppresses a hard-bounced address at once, and a soft-bounced address after 3 soft bounces. A sudden jump in bounces at one provider is often a block, not bad addresses. See email bounce codes and what to do with each.
Results by recipient domain
Totals hide problems. If Gmail readers click as usual and Microsoft readers stop clicking, your overall numbers dip a little while one provider is filtering you. The campaign report in EmailMarketer.io breaks results down by recipient domain, so compare each large provider with its own past sends. For the provider’s own view, read Google Postmaster Tools and Microsoft SNDS. Senders of 500,000 emails a month or more also get a named deliverability contact who reads these numbers with them. See how we handle deliverability.
Where opens still help
Opens are not useless. They are a biased count, and the bias is roughly stable for the same list over a short time. Use them to:
- Compare two subject lines sent to random halves of the same list on the same day.
- Spot a sudden drop at one provider, which can point to filtering.
- Compare this week’s newsletter with last week’s.
Do not use opens to:
- Report reach or readership to anyone outside your team.
- Compare your list with industry benchmarks.
- Decide alone who is inactive and gets removed.
Build engagement segments on clicks
Sunset and re-engagement rules decide who keeps getting your email, so they need the most reliable signal you have. EmailMarketer.io segments can filter on opened, not opened, clicked, not clicked, bounced or complained, each with a time window.
A rule for a weekly newsletter might be:
- Engaged: clicked in the last 90 days.
- At risk: no click in 90 days, but opened in the last 90 days. They may be reading on Apple Mail, or the opens may be automatic. Keep sending, but less often.
- Inactive: no click and no open in 180 days. Send a re-engagement email, then stop.
The windows depend on your frequency. A daily newsletter can use shorter ones than a monthly product update. Our sunset policy guide covers the full process.
Decision table
| Question | Use |
|---|---|
| Did this campaign work? | Conversions and revenue, then main-link clicks |
| Which subject line was better? | Opens or clicks from an A/B split sent at the same time |
| Is a provider filtering us? | Results by recipient domain, bounces, Postmaster Tools |
| Is our list getting tired of us? | Unsubscribes and complaints per send, over time |
| Who should we stop emailing? | No clicks over a long window, plus no opens |
| How many people read it? | You can’t know. Report clicks and conversions instead |
The reports in EmailMarketer.io show results by recipient domain, bounce and complaint rates per campaign, and conversions and revenue once you track them. See reports.