The best time to send an email, and why the studies disagree

What every send-time study is actually ranking

A send-time study works by taking a large pool of campaigns, bucketing them by the hour they were sent, and comparing open rates between the buckets. Whatever the headline says, the finding is always of the form: messages sent in this bucket recorded more opens than messages sent in that one. Everything else — the recommended hour, the heatmap, the weekday ranking — is that comparison with a story attached.

Two problems live inside that method even before the measurement is questioned. The first is that the pool is whoever uses one platform, which means the result describes an industry mix rather than your audience; a study dominated by retail promotions is not evidence about a teaching sequence. The second is that senders are not randomly assigned to hours. Experienced senders, who have better lists, choose their send times deliberately, so the best-performing hour in an observational dataset is partly a measure of which hour careful people prefer.

Neither of those is fatal on its own. What is fatal is that the number being compared stopped meaning what it used to mean, and every study published since has gone on comparing it anyway.

Why opens stopped being a measurement in 2021

An open is not an event a mailbox reports. It is inferred from a one-pixel image loaded from the sender’s server when a message is displayed, and the whole method depends on the image being fetched at the moment a person looks at the message. In 2021 Apple shipped Mail Privacy Protection, which, for readers who turn it on, routes remote content through a proxy and preloads it whether or not the message is ever opened. Apple’s own documentation describes it plainly: it hides the reader’s IP address and prevents senders from seeing whether the mail has been opened.

The consequence is not that opens went down. It is that they became a mixture of two things that cannot be separated afterwards — people who read the message, and software that fetched a pixel on their behalf — with the proportions varying by audience, by device, by client and by month. A rate computed from that mixture can still be printed to one decimal place, which is exactly why it is still printed.

Nobody can tell you how large the proxy share is for your list, and this page is not going to invent a figure for it. Apple publishes none. The third-party estimates of email client market share move every quarter, so a number copied into a guide is wrong within one. What can be said without a number is the part that matters: any hour-by-hour comparison of open rates is comparing a quantity whose composition changes between the buckets, and an average of two different things is not evidence about either.

The question that does have an answer: whose clock

Timing questions divide into three, and only one of them is usually asked. Which hour, which day, and in whose time zone. The third is the one with a real answer and the one most often decided by accident, because the default in most tools is the sender’s clock, or the server’s, and a server clock is generally UTC.

Send in the reader’s local time. A message sent at nine in the morning your time reaches a reader eight hours west at one in the morning, which is not a subtle penalty: it arrives at the bottom of a pile that accumulates overnight, and by the time that person looks at their mail it is under everything that arrived after it. Where local time is genuinely unknown, the least bad default is the time zone most of the list is in, stated as a decision rather than inherited from a server.

That is also the only timing choice with a mechanism behind it rather than a correlation. Nobody has to believe a study to accept that a message which arrives while somebody is asleep is read later, further down, alongside more competition.

The hour matters much less than never moving it

Within the waking day, the difference between plausible hours is small and unstable, and it is dwarfed by an effect that costs nothing to obtain: arriving at a predictable time. A reader who receives something at the same hour each morning acquires a habit around it, and a habit is worth far more than an optimisation of a rate nobody can measure properly. The conventional window is somewhere between six and nine in the morning, which works mostly because it puts the message near the top of the pile a person opens first.

For a sequence this compounds. The second message benefits from the first one having arrived at a recognisable time; the third benefits from the second; by the fourth, the arrival itself is the reminder. Move the hour around in search of a better slot and you trade that for noise, because the variation you are chasing is smaller than the variation between weeks.

The one exception is the first message. When somebody has just asked for something, the best moment to deliver it is immediately, while they are still looking at the inbox they clicked a link in. Holding a welcome or a first lesson until tomorrow’s send window costs recognition, and recognition is the entire asset.

For a course, the weekday question dissolves entirely

Best-day-of-the-week advice assumes a broadcast: one message, everybody at once, one calendar day to choose. A sequence counted from each person’s own start does not work that way. Somebody who confirms on a Thursday gets day two on Friday and day five on Monday; somebody who confirms on Saturday gets an entirely different overlay of the sequence onto the week. There is no single weekday to optimise, because every subscriber is on their own.

This has a pleasant side effect for deliverability. A broadcast concentrates a month of volume into one hour, which is the pattern that looks least like ordinary mail to a receiving provider; a sequence fans out one message per subscriber per day, spread across whichever days people happened to sign up. The volume smooths itself without anybody configuring a throttle.

What replaces the weekday question is a cadence question, and that one is worth thinking about. One message a day is what a five-day course promises, and it is the cadence a reader agreed to when they signed up. Skipping a day to avoid a weekend breaks the promise more expensively than a weekend send breaks anything, because the reader counted.

What to measure instead, now that opens are unusable

Every signal that survives has one property in common: it requires a person to do something deliberate. A proxy can fetch an image; it cannot decide to click a link, write a reply, or unsubscribe. So the signals worth keeping are the ones with an intention behind them, and the useful comparison is between your own sends rather than against an industry benchmark computed from somebody else’s mixture.

Compare like with like, and over enough sends to mean something. If you want to test a send hour, change one thing, hold it for several cycles, and read a signal from the table below rather than an open rate. Most people discover that the effect they were chasing is smaller than the week-to-week variation of their own list — which is itself a useful finding, because it frees up the attention for the subject line and the first sentence, where the differences are large.