Are open rates accurate?

If you've ever celebrated a 60% open rate, there's a good chance you were partly celebrating your subscribers' email clients doing automated housekeeping. Open rates have never been perfectly accurate, but right now they're more noise than signal for a lot of senders.

Here's how open tracking actually works. Your ESP embeds a tiny invisible image (a tracking pixel) in each email. When that image loads, it counts as an open. The problem is that image can load without a human ever reading your email. Security scanners fetch it to check for malicious content. Corporate gateways pre-load it to scan for threats. And since September 2021, Apple Mail's Mail Privacy Protection (MPP) pre-fetches every image in every email, automatically, before a subscriber even opens the message.

That last one is the big one. Litmus tracks opens across its user base, and the share triggered by Apple's Mail Privacy Protection went from 14.35% in October 2021 to 52.22% in May 2022, 53.95% in February 2023 and 54.25% in February 2024. One before and after look, from Omeda in 2021, saw unique open rates go from 15.2% to 19.1% once MPP arrived. That means about 1 in 5 reported opens were new ones that had not been there before. Omeda's senders are mostly B2B media, and the two periods were different sends, so treat it as a picture, not a rule. If even half your Apple Mail recipients trigger a phantom open, your reported open rate can be inflated by 20 percentage points or more. Some senders saw their open rates jump about 5 points overnight when MPP rolled out, others as much as 40, the most on Apple-heavy lists. That's not engagement. That's automation.

And it's not just Apple. Outlook and Google Workspace both have image pre-fetching or security scanning behavior that can trigger pixels, too. The scale is smaller, but it adds up. SparkPost (now Bird) found in December 2021 that Gmail's image prefetching made up 1 to 6% of Gmail open events, worth up to 2 points of open rate.

So what can you actually trust?

• Clicks are a much stronger signal. A human has to actually interact with a link for that to fire (though bots do click, too, so watch for patterns like instant clicks from the same IP).
• Trend direction still matters. If your open rate drops sharply month over month with no obvious change in list composition, something real has likely shifted, even if the absolute numbers are off.
• Relative comparisons are still useful. Comparing segment A to segment B, or subject line A to subject line B, is valid as long as both are affected by the same noise sources in roughly equal measure.

What you shouldn't do is set an "open rate threshold" to trigger re-engagement campaigns or suppress contacts. If someone is being counted as an opener purely because MPP pre-fetched your pixel, suppressing them based on non-opens, or keeping them active based on phantom opens, both lead to bad list decisions.

The honest answer is that open rates still have some use, just not as a precise metric. Treat them like a barometer. They tell you if pressure is rising or falling, but they won't tell you the exact temperature.

If you want metrics that actually reflect engagement, clicks, replies, conversions, and unsubscribe rates are all closer to the truth. Those are worth building your decisions around.