Open rate

Open rate is the share of delivered emails recorded as having been opened. It is the number most email platforms put at the top of the report, and it is the number people quote when they say a send went well. It is also, since 2021, a measurement of something other than what its name describes.

In short

• An open is recorded when a hidden image loads, nothing more.
• Apple Mail loads that image for people who never opened the message.
• Clicks measured against delivered messages are the honest replacement.
• It still detects a delivery failure, which is worth keeping it for.

How an open is counted

Everything strange about this metric follows from the mechanism, so it is worth knowing exactly what happens.

Every marketing email carries a tracking image about one pixel across, invisible against the background. The image lives on the sending platform’s server. When a mail app displays the message and fetches that image, the platform notes that the message with that address attached has been opened.

That is the entire method. Email has no channel back to the sender and no way of knowing whether a human was present. Two consequences follow immediately, and they point in opposite directions. Someone who reads every word in an app that blocks images is never counted at all. Anything that loads images without a person being involved is counted as a reader.

What Apple changed

The second of those consequences stopped being an edge case in 2021.

Apple introduced a privacy feature in Mail that recipients can switch on, and many did. It routes messages through Apple’s own servers and loads the remote images there, in advance, regardless of whether the person ever opens the message. The sender’s tracking pixel is one of those images. The open is recorded, timed to the moment Apple fetched it, from a location that is not the recipient’s.

So opens now arrive for people who did not open anything. How badly this affects your particular number depends on how much of your list reads mail on an iPhone, which is not something you chose and not something you can see. Two businesses sending identical emails to lists of the same size can report open rates that are not comparable, and neither of them is wrong.

There is a quieter casualty. Several platforms offer to send each person at the hour they usually open, and that recommendation is built from open timestamps. When a proportion of those timestamps record the moment a server fetched an image overnight, the best time to send is being calculated partly from machines. The feature is not useless, but it deserves less trust than the interface implies.

What to read instead

The replacement is not a new metric. It is one that was always better and was overshadowed by an easier one.

A click-through rate measured against messages delivered describes something a person decided to do. No privacy proxy manufactures a click, because clicking means loading a page and continuing a session. Below that sits the only measure that answers the question anybody actually asked, which is what those visitors went on to do: the conversion rate of the traffic, tracked by tagging the links so the visits are identifiable in analytics rather than filed as direct.

Click-to-open rate deserves a specific warning, because it is still shown prominently. It divides clicks by opens, and its denominator is now partly fictional, so it moves in response to your audience’s choice of phone.

The one thing it is still good for

Being unreliable is not the same as being empty, and there is a use worth keeping.

A sudden collapse is real information. Open rate does not usually fall by half because a subject line was poor, and rarely because the from name changed, though that is the one change worth ruling out first. It falls by half because messages stopped arriving in inboxes, and that makes it a rough early warning for a deliverability problem, which is worth catching within days rather than months.

The trap sits inside your own platform. Most email tools let you build a segment from people who opened recently, and that segment is now contaminated with addresses belonging to people who have never looked at anything you sent. If you use engagement to decide who still deserves mail, rebuild it around clicks. The audience gets smaller and stops lying to you.

Words you will hear

• Unique opens. One count per recipient, not one per image load, which is the figure most reports show by default.
• Total opens. Every load counted, including the same person twice.
• Tracking pixel. The hidden image doing the measuring.
• Machine opens. Opens generated by a proxy or a security scanner rather than by a person.
• Click-to-open rate. Clicks divided by opens, now built on a denominator that cannot be trusted.
• Engaged segment. A slice of the list defined by recent activity, and worth defining by clicks instead of opens.