Email open rates are the first number most people look at. They are also the least trustworthy number in the report. An open is not a read. It is a record that a mail client fetched a small image, and that event has only a loose link to a person paying attention.
That does not make the figure useless. It makes it a rough signal that needs company. So this article covers where the number comes from, what it can honestly support, and which decisions it should never drive on its own.
Where email open rates come from
Open tracking rests on a picture. The platform hides a tiny transparent image inside the message, and the mail client fetches that image when it renders the mail. Nothing else in the message reports back.
So the measurement rests on one thing outside your control, which is whether the reader's software loads pictures. Some clients load them by default. Others hold them back until the reader asks. A few never load them at all.
Plain text messages carry no image, so they produce no opens. Somebody who replies to your email from a text only client counts as a person who never opened it.
Sent, delivered and opened mean three different things
Open rate usually means recorded opens divided by delivered messages. Delivered means the receiving server accepted the message, which is not the same as reaching an inbox. A message sitting in a spam folder still counts as delivered.
Some tools divide by sent instead. Others count unique opens where their neighbour counts every load. Both versions wear the same label on the dashboard, so check which definition sits behind a number before you compare it with anything.
What email open rates can honestly tell you
The figure earns its place as a trend line rather than as a fact. Watch the same list, the same sending domain and the same style of campaign over several months. A steady line means nothing changed. A sharp fall deserves a look.
Three uses stand up reasonably well. First, email open rates work as an early warning for delivery trouble. Second, they show whether a list still has a pulse. Third, they give a rough read on subject line tests, so long as you treat small gaps as noise.
None of those three is a claim about how many people read the message. Each one compares the same measurement against itself over time.
What email open rates cannot tell you
They cannot tell you that somebody read anything. A campaign can post a healthy figure and sell nothing at all, because reading and buying are separate acts.
Nor can the figure support a comparison with anybody else. Two lists differ in mail clients, in where the addresses came from, and in how much machine traffic they attract. A number that looks poor next to somebody else's may sit on the healthier list of the two.
Least of all does an open describe one person. It records what a piece of software did, not what a reader decided. So any judgement about a single subscriber that rests on an open will sometimes be wrong.
What mail privacy features did to email open rates
Some mail providers now load images through their own servers rather than the reader's device. The stated reason is privacy, and it works, because the sender learns much less about the person. It also means the image can arrive before anybody opens the message, or without anybody opening it at all.
Two things follow wherever that happens. Recorded opens rise for reasons that have nothing to do with interest. The detail that used to ride along with an open turns into fiction too, including the time of day and the rough location of the reader.
Meanwhile the opposite fault never went away. Readers who block images stay invisible, so genuine attention goes unrecorded. Email open rates now carry an inflated count and a missing count at the same time.
Which providers behave this way, and how, is not something to take on trust from an article. Read what your own tool reports about mail clients instead, since that is the only version of the story that describes your list.
Machine opens and what they do to a report
Security scanners open mail before a person sees it. Corporate filters follow links to check where they lead. A preview pane can load the images too, even when the reader deletes the message a second later.
None of that is fraud. It is machinery doing its job, and it lands in the same column as a human being. Corporate mailboxes usually sit behind more of that machinery than personal ones, so a business list tends to carry more of it.
How to tell how much of your count is machinery
You cannot separate the two cleanly. You can still get a direction of travel, and three checks do most of the work.
First, put the open line and the click line on one chart. Real interest tends to move both. When opens climb while clicks sit flat, the counting changed rather than the audience.
Second, look at when the opens land. A cluster of them recorded in the first seconds after a send is machinery, not an eager reader. Few humans are that quick.
Third, split the figures by mail client if your tool allows it. One client jumping on its own points at a software change. A broad move across all of them points at your list or your sending instead.
None of this hands you a clean number, and any tool promising one is guessing. Treat the answer as a direction, then judge the campaign on clicks and orders anyway.
Reading a fall in email open rates
A drop is a question rather than a diagnosis. Work through the causes in order, because the cheapest explanations are also the most common ones.
- Tracking changed. Somebody switched tools, moved domain or turned image tracking off.
- Client mix changed. A large group of readers arrived on software that behaves differently.
- Delivery slipped. More mail is landing in spam, which no subject line will fix.
- The list aged. Old addresses stop reading long before they unsubscribe.
- Content drifted. Send the same offer every week and people stop bothering.
Only the third of those is an emergency. If mail is landing in spam, opens fall because nobody sees the message. The fix then belongs in deliverability rather than in the copy.
Numbers that deserve more weight than email open rates
Every signal below records something a person chose to do. That is the difference, and it is the whole reason each one beats an open.
| Signal | What it proves on its own |
|---|---|
| Open | Only that software fetched an image |
| Click | That a person acted, although not that they wanted to buy |
| Reply | That a person spent real effort on you |
| Order or enquiry | That the email did its job |
| Spam complaint | That somebody wants you gone |
Notice that email open rates sit at the bottom of that list. They still belong in the report, because a collapse tells you something real. If you want the fuller version, we set out the email metrics worth reporting and the order to put them in.
A worked example of how the arithmetic misleads
A worked example, with round numbers picked to make the sum easy rather than to describe a typical campaign. Send 1000 emails, record 300 opens, and the open rate reads 300 in every 1000. Now suppose 100 of those opens came from scanners and image proxies. Real human opens are 200, so the honest rate is 200 in every 1000. The campaign did not change at all. Only the counting did.
That is the trap in one sum. The number moved without the audience moving, and no amount of rewriting would have caused it.
How to report email open rates without misleading anyone
Label the figure for what it is. Write recorded opens rather than people who read this, because the report will outlive whoever wrote it.
Then show a trend rather than a single campaign, since one send settles almost nothing on its own. Put clicks, replies and orders above it on the page, so the eye lands first on the numbers that record a decision.
Finally, say out loud what the number cannot do. One line in a footnote saves an argument later, when somebody points at a rise and calls it success.
Automations that should never run on an open
Any flow that treats an open as interest now misfires. Re-engagement is the obvious victim, because a machine open marks a dead address as alive and keeps it on the list.
So define engagement with actions instead. A click, a reply, a site visit, an order or a preference change all mean somebody moved. Those signals also make email segmentation honest, since the segments then describe behaviour rather than image loading.
What to do with email open rates this month
Nothing here says drop the measure. It says demote it. An open works as a smoke alarm and fails as a thermometer, so give it the job it can actually hold.
- Find out how your tool defines an open, then write that definition at the top of the report.
- Rebuild the report so clicks, replies and orders sit above email open rates.
- Rewrite any automation that keys on an open, so it keys on a click or a reply instead.
After that, leave the figure alone for a quarter and watch the shape of the line. A run of campaigns, read next to clicks and orders, starts to be worth the space it takes.
One question sits underneath all of this, which is what the list is for in the first place. Our plain guide to email marketing works through that, and the answer decides which numbers belong at the top of the page. Once the goal is an order or an enquiry, steering by opens stops looking sensible.
Frequently asked questions
What is a good email open rate?
There is no honest benchmark to hand you, and any figure quoted without the list behind it is guesswork. Counting depends on which mail clients your readers use, whether their software loads images, and how much machine traffic hits your mail. Two lists in the same trade can record very different numbers while performing identically. So compare your list against itself over time, and judge a campaign on clicks, replies and orders instead.
Why did my open rate suddenly jump?
Check the boring explanations first. A jump usually follows a change in tracking, a change in the mix of mail clients on your list, or a batch of new addresses from a source that attracts scanners. Machine opens can arrive in a lump and lift the line overnight. If the click count and the order count stayed flat while opens rose, then almost nothing real happened and the counting changed instead.
Do opens still work for subject line testing?
They work as a rough guide and nothing more. A test judged on opens now mixes real readers with proxies and scanners, so small gaps between two versions mean very little. Treat a large, repeated gap as a signal and ignore the rest. Better still, score subject lines on clicks rather than on email open rates, because a line that wins attention and loses the click has not helped you.
Should I remove people who never open my emails?
Not on opens alone, because a reader who blocks images looks identical to a dead address. Build your inactive segment from actions instead: no clicks, no replies, no visits and no orders across a long window. Then try a small win back campaign before you delete anybody. Clearing out quiet addresses usually helps delivery, so the job is worth doing, as long as the definition of quiet never rests on a tracking pixel.
What is the difference between unique and total opens?
Unique opens count each recipient once, however many times they open. Total opens count every load of the tracking image, so one person reading three times adds three. Your dashboard may report either, and the label on it rarely says which. Check before you compare a campaign against last month, because switching between the two moves email open rates without anything changing in the mail itself.
Can I stop counting opens altogether?
Yes, and some senders do exactly that. Turning the pixel off removes a noisy figure, although it also removes your earliest warning that delivery has gone wrong. A middle path usually works better: keep counting, demote the number in every report, and never let an automation depend on it. If you do switch tracking off, tell whoever reads the report, otherwise a missing line looks like a collapse.
Why do email open rates differ between two tools?
Because the tools count different things. One may report unique opens while another reports total opens, and each divides by a slightly different denominator, such as sent, delivered or accepted. Filtering may differ too, since a platform can strip known machine opens or leave them in place. Email open rates are therefore a house measure rather than an industry one. Pick a single tool for decisions, keep the other as a cross check, and stop trying to reconcile them exactly.
What should I report instead of email open rates?
Lead with the numbers that record a decision: clicks, replies, enquiries, orders and revenue where you can attribute it. Add unsubscribes and spam complaints, because those are the honest cost of sending. Keep email open rates on the page as a trend line, labelled as recorded opens rather than as readers. A board that sees the money first stops asking why one campaign beat another on a measure nobody controls.