What your open rate is actually measuring

You sent a campaign on Tuesday and the dashboard says 61% opened it. That feels like a report card, and for a moment it is a good one. But the number is answering a narrower question than the one you asked. You wanted to know how many people read your email. What you were handed is a count of how many times an image inside it was fetched — sometimes by a person, often by software that has never read a word of anything.

An open is recorded when a tiny image embedded in the message gets loaded. For years that was a rough proxy for a human opening the message, and it worked well enough that the whole industry built habits on it. Then mail apps started preloading images for every message on arrival, whether or not the recipient ever glances at it, and corporate security tools started fetching everything in an email before it is delivered so they can scan it. On the other side, plenty of people still read with images off, and they register as never having opened anything at all. The result is a number that is inflated for one part of your list and blind to another, in a ratio you cannot see.

The arithmetic, made up but honest

Say 10,000 emails were delivered and the tool reports 6,100 opens, so 61%. Now suppose — hypothetically — that 45% of that list reads mail in an app that preloads images. Those 4,500 people register as opens on arrival. The other 5,500 produced the remaining 1,600, which is 29%. So which is your campaign's open rate, 61% or 29%? Neither, exactly. Both are made of guesses about a split nobody reports to you. That is the part worth sitting with: the precision on that number is theatre.

an open is a machine's curiosity, not a person's

It is not useless, it is just the wrong job

Open rate still works as an alarm. Compare the same list to itself across a few weeks and the machine noise stays roughly constant, so a sudden drop of twenty points is real information — usually about delivery, not about your writing. Landing in spam looks exactly like that. So does a change in how you send. Keep the number, keep watching the shape of it, and stop treating the absolute value as a grade.

What it cannot do is settle an argument about content. When the difference you are looking for is three points and half the readings are automatic, three points is well inside the noise. This is the quiet reason so many subject line tests feel inconclusive: they are inconclusive. Every test needs a gap big enough to see past the noise, and this particular number comes pre-loaded with a lot of it. If your list is small, you already knew tests were hard here — the auto-opens make it harder still.

Clicks are better, and also not clean

The natural next move is to judge on clicks, and that is a real improvement, because a click is a deliberate act. Mostly. The same security scanners that fetch images also follow links to check where they go, which means a share of your clicks arrive from something that will never buy anything. You can usually spot it: clicks that never turn into a visit on the other end, arriving within seconds of the send, all from a handful of your business subscribers.

So compare the two sides. Clicks reported by the email tool, versus sessions the landing page actually saw from that campaign. If the first is much larger than the second, the gap is ghosts, and your real click rate is the smaller number. It is one of the few checks that costs you five minutes and changes what you believe.

Promote the thing that happened after

The most honest number in an email campaign is the count of things that happened because of it: trials started, replies written, purchases made, calls booked. They are small numbers, and small numbers feel less satisfying than 61%, but they are made of people. Nine signups from a send is a fact. A 61% open rate is an estimate of an estimate.

For a small list, count events rather than rates. Rates on 400 recipients swing wildly for reasons that have nothing to do with you, which is the same trap as reading a one-day dip in traffic as a trend. Counts do not pretend to a precision they do not have. Write down: this send produced four trials and one reply. Next month, five and three. That is a series you can steer by.

Two more signals are worth more than they get credit for, because humans have to work to produce them. Unsubscribes are a deliberate act, and a rising unsubscribe rate at a steady open rate is a clean message about cadence or relevance — the list is not broken, the rhythm is. Replies are the other one. Nobody's mail app writes you a reply by accident.

What we show you, and why

When we pull email numbers from a connected tool, we show what that tool reports, including open rate, because hiding it would just send you to another tab. What we try not to do is dress it up. If you are building the one number you check each morning, build it out of the thing that happened after the click — signups, purchases, whatever your goal actually is — and let opens sit further down the page as a health check.

Here is the rule of thumb we would keep if we could only keep one: judge a send by what it caused, not by what it was reported to have reached. Then ask yourself the uncomfortable version of the question. If your open rate dropped ten points next month but your signups from email doubled, would you call that campaign a failure? If the answer is no, you already know which number was never really running the show.