Late last year, MailChimp rolled out our Send Time Optimization feature. Built on the Email Genome Project, which powers many of our other email marketing products, Send Time Optimization does just what you’d think: It suggests to users when they should send their campaigns.
"So," you might be thinking, "when is the optimal time to send?" Well, it depends on who you are and who you’re sending to: Since the Email Genome Project stores email address engagement data for billions of addresses at the individual level, these send time recommendations are personalized per user based on the subscribers on their list.
But that doesn’t mean we can’t tackle the question at a high level anyway. In this post, we’ll look at aggregate patterns from the send time optimization system and then investigate some drivers behind why different lists often have different optimal send times.
A note on how to read the graphs in this post: For this analysis we calculated the best time to send to individual addresses in their local time zones. "Best time to send" is a calculation I explain here (it’s a function both of engagement and volume). The graphs then track the percentage of email addresses out of the population that have a particular time as their optimum. For example, 1% of email addresses might have 3am as their optimum, while 6% have 9am. Across the time horizon of a graph, these percentages will sum to 100%, because everyone has some time that’s optimal. (Also, these names can get unwieldy, so for this post I’m calling them STO and EGP. Cool?)
The easy answers
For all the readers who want the high-level, easy answers about when to send, here I’ll lay out what EGP’s several billion email addresses have to say. I’ve separated time of day and day of week to make the results easier to read, but even when merged into single week graphs, the results are the same.
To start, let’s look at day of week effects. STO’s data shows us that, in general, it’s most optimal to send to most subscribers on a weekday. Sunday is the best day to send to the fewest number of subscribers, while Monday through Friday are all quite similar in terms of the percentage of email addresses that have that day as optimal.
Note that no single day wins hands-down. This is what we should expect when studying the data of billions of humans’ inbox activity.
Now, while the weekend is clearly less preferable compared to the work week, that does shift slightly depending on what type of content you send (since that affects the subscribers you have).
For example, business-related content has a less than typical number of subscribers with weekend optima. On the flip side, recreational content usually enjoys a healthier than typical percentage of subscribers with weekend send time optima. Note though that in either case, these percentages are still lower than 50%, implying that no matter what content it being sent, most lists are comprised of a majority of recipients for whom the optimal send time is during the work week.
Now, let’s zoom into the hourly level.
When we look at the typical distribution of optimal send times across MailChimp’s whole system, the peak is at 10AM in the recipients’ own time zones.
Unless you have good reason, I’d avoid sending your content too early (say, 3 or 4 am) in the recipients’ time zones. If you’re on the east coast, but many of your subscribers are in California, consider using delivery by time zone on your morning newsletter to make sure you don’t send to those west coasters too early.
Note that the 10am peak is less than 7%. That means no matter what time you choose to send to a list, the majority of subscribers on your list have some other time that’s optimal for them. There’s no one time during the day when everyone (or even half of everyone) drops what they’re doing and says, "Now is the time, and this is the place, to engage with email." That’s just not going to happen, and I love that the data bears this out.
When we break the high-level time of day curve into some content types, we see that 10am time actually fracture. For instance, content in the hobbies category in MailChimp goes to subscribers whose optimal send time peak is earlier, around 8am. Perhaps these readers check a lot of this content before they go to work.
Note that retail and hobby content have many more subscribers with optimal send times after business hours (not enough to shift the peak from the morning, though).
Not content, but audience
We’ve got our easy answers out of the way (send during the work week in midmorning), and we’ve seen how these answers shift a bit in terms of weekend preference and evening preference based on content type. These content-type dependencies are the first indication that our monolithic answers are disguising more nuanced truths deeper in the data.
When you think about it, is the type of content you’re receiving the main driver of when you open and read it? For me, that’s not true. I’ll read retail stuff at work (sorry, boss!), and I’ll read data science content while I’m at home on the couch (sorry, lovely wife!).
The primary driver of when I engage with email (donning my Captain Obvious hat right now) is my life. A lot of the analysis and anecdotal evidence around send time stops at business type, but in order to understand when to send to your audience, you need to understand your audience, not your content. Who are these folks? What stage of life are they in? When do they have a free moment to scroll through promotional material on their phone?
To illustrate a quick point, let’s take a look at how the distribution of optimal send times for Republicans differs from that of Democrats.
There’s no statistically significant difference whatsoever. Why is that? Because political affiliation at a high level has little to do with what drives a person’s schedule and their ability to check email. Large amounts of data don’t have to overturn common sense hypotheses. Common sense would tell us that political affiliation is too broad and tangential a category with respect to send time optimization.
So let’s think a moment about what factors would have an effect on when people are available and ready to engage with email.
How about the location and culture in which they live? After all, people in different countries have different schedules. Let’s look at send time optimization split out into a few countries:
A much larger percentage of folks in Spain engage with email in the 10 and 11am hours, while Norway—which currently has a very long summer day—plateaus out with plenty of folks having optimal send times well after business hours. Egypt, on the other hand, has a much higher percentage of recipients whose optimal send time is in the predawn hours.
So location matters. What about age?
College-aged recipients have an optimal send time distribution that is shifted to the right with a peak at 1pm instead of the typical 10am. Given the sleep patterns of college students, especially in the summer months, this comes as no surprise.
So age is certainly a step in the right direction. But why don’t 30 year olds sleep til noon like college kids do? In my case, it’s because my kids would never let me, but for the most part, people have to go to work!
What about occupation, then?
When we look at occupation, we see some of the largest differences. In the graph below, I’ve pulled optimal send time distributions for recipients who are lawyers, bartenders, and neonatal nurses.
Bartenders, like college students, get a later start on the day, and their whole distribution shifts right, giving them a peak at around 1pm. Lawyers, on the other hand, have a highly condensed distribution, with two peaks: first thing in the workday at 9am, and then again after lunch. Very few lawyers have optimal send times that are outside of work hours.
As for neonatal nursers, their workdays are all over the place and so their distribution gets widened and is less peaky. Indeed, many of these nurses have optimal send times well past 8pm.
So what if I’m not sending to Norwegian bartenders?
The graphs above illustrate the nuance in determining the optimal time to send to a list. There may be one high-level easy answer, but it falls apart a bit for certain audiences. Luckily, MailChimp’s STO system captures all that nuance, so feel free to try out its suggestion for your list.
Furthermore, none of these above results are surprising. The fact that bartenders don’t want to engage with email at 8am is pretty obvious, if you think about it. And that’s a good thing! It means that STO, while backed by insane amounts of data and math, isn’t magic. You can formulate your own hypotheses based on what you know about your audience and then A/B test those send time hypotheses. (Remember, A/B tests on send time average a 22% improvement in engagement.)
So I hope this post helps put the age old "When do I send?" question to bed. Just remember:
- Send midmorning during the work week…
- …unless you know something about your audience, which you really should. In which case, the previous point is probably wrong.
- Use STO or A/B testing to get a better answer.