Machine opens, and why your open rate moved

Since 2021, Apple Mail Privacy Protection has fetched every email its users receive the moment it arrives, whether or not the person ever reads it — and Apple Mail is where a large share of consumer email is read. Corporate security gateways do the same thing for a different reason: they load every image and follow every link to check what is behind it. To a tracking pixel, both look exactly like an engaged reader.

This is why open rates rose industry-wide without anybody getting better at writing subject lines, and why a tool that quotes an unfiltered open rate is quoting a number about machines as much as about people.

What doTask does. As each open and click arrives, it is judged. A fetch is a machine when the browser it names is a known proxy or scanner, when it arrives within ten seconds of the send (nobody reads that fast — that is the mailbox opening it on arrival), or, for clicks, when one recipient takes three different links inside half a minute, which is a scanner checking them rather than somebody reading.

Nothing is deleted. The open is still recorded, the click row is still there, and the person still shows as having opened on the recipient list. What changes is which of those numbers the reported rates count. Under the headline on every send is a line naming how many machine opens and scanner clicks were left out, so a lower number than you are used to explains itself rather than looking like a fault.

The raw number has not gone anywhere: in Analytics → Reports the Email sends dataset carries Open rate, unfiltered alongside Open rate, and Machine opens as a rate of its own. A machine-open share of 40–60% is ordinary; it is worth watching only as a trend.

Two places this quietly makes better decisions. An A/B test is judged on people — Apple opens both versions equally, which drags a real difference toward no difference and would leave a test saying “not enough yet” for ever. And whether somebody counts as unengaged, and what hour of day doTask thinks they read at, now follow their real opens, so a send that skips the unengaged skips the right people.

Clicks were always the sounder measure and they still are, machine sweeps aside. When opens and clicks disagree about which send did better, believe the clicks — and believe the revenue over both.