Email open rate:
the sum, the denominator and the privacy catch

How an open gets recorded in the first place

An email has no built-in way to tell the sender it was read. Marketing platforms add a one-pixel image to each message, with a unique address. When the mail app loads that image, the platform logs an open for that recipient.

That mechanism explains almost every oddity in open data. If images are blocked, a read email shows no open. If an app loads images in advance, an unread email shows one. The metric measures image loads, which usually, but not always, correspond to a person reading.

A related quirk is the open that happens twice. Forwarding an email, opening it on a laptop and again on a phone, or having a mail provider re-fetch the image can all log extra events. Counting unique recipients, not events, protects you from most of this.

Because of these quirks, some teams now treat opens as a deliverability and trend signal only, and set their goals on clicks, replies and conversions instead.

The open rate formula

Use unique opens, one per recipient, for the headline rate. Total opens count the same person reopening a message on several devices and can push the percentage past anything meaningful.

Worked example: bounces change the base

The 0.8 point difference came from nothing but the choice of denominator. When comparing with another report, make sure both used delivered emails or both used sent emails.

Rounding conventions also differ. One platform may truncate 19.23% to 19%, another keeps one decimal. Pull the raw counts, which are exact, and do the division yourself when two systems disagree.

For an automated sequence, report the rate per step rather than for the sequence as a whole. A welcome email routinely earns two or three times the opens of the fifth message in a drip, and an average of the two says little about either.

Why opens became less trustworthy

Since 2021, mail apps on some devices have offered a privacy setting that downloads images for every message in the background, whether or not the recipient looks at it. Those downloads register as opens, usually at the moment of delivery, so a list with many such readers shows a higher open rate with no real change in attention.

The opposite distortion exists too. Plain-text emails carry no image, and some corporate clients block images by default. Each one is a genuine read that never appears. The two errors do not cancel in any predictable way, which is why a single campaign figure is a soft number.

How many opens are enough to trust a difference

Open rates move around by chance even when nothing changes. On a send of 1,000 delivered emails at a true 20% rate, the observed figure will often land anywhere between roughly 17.5% and 22.5% purely through sampling variation. On 10,000 the same spread narrows to about 19% to 21%.

That is why a subject-line test on a small segment needs a large gap to be believable. A 2-point difference over 500 recipients per version is noise. The same 2 points over 5,000 per version is worth acting on. When in doubt, run the test again and see whether the winner repeats.

Levers that actually move the figure

Deliverability sits underneath all of these. If a large share of your messages lands in spam or in a promotions tab, the delivered count looks fine but nobody sees the email, and no subject line will fix that.

Frequency is another lever people overlook. Sending twice a week to a list that only expects a monthly note can raise short-term opens and then collapse them as readers tune out. Watch unsubscribes and complaints alongside opens before deciding that more mail is working.

Reading your own report sensibly

• Segment by recent subscribers against older ones; an old list drags the average down while the engaged part is healthy.
• Track opens for a fixed window, such as the first 48 hours, so campaigns are comparable.
• Keep a record of sample size. On a list of 500, a 3-point change is 15 people.
• Pair open rate with unsubscribe and complaint rates to see whether higher opens came at a cost.

A calculator takes the sent, bounced and opened counts and returns the rate on either base, which makes it a useful sanity check against what the platform shows.

Putting the number next to other signals

Open rate remains useful as an early warning. A steady fall across several sends, with clicks falling too, suggests a deliverability problem or a tired list. A fall in opens with stable clicks is more likely to be a change in how the mail apps count.