There’s a lot of bad advice out there.
For years, marketing consultants, influencers, and vendors alike have claimed that there are certain times during the day in which to send out emails for maximum response.
9 a.m. Tuesdays! If you’re in technology, 10 a.m. Wednesdays! Never send that on a weekend!
It’s largely hogwash. Think about how you treat your own inbox. I prefer to do a quick cleanup and scan very early in the morning almost every day (including weekends), another cleanout around mid-afternoon (when I’m annoyed with the volume), and sometimes a final one in the evening (weekdays only).
You might be different. And that’s totally unsurprising. So why is it that we’re constantly told there’s a single best send slot for email marketing?
Two reasons: 1) Marketing Automation platforms were designed to only give you one candidate send slot for any given population and 2) most of these platforms aren’t smart enough to take advantage of the audience preference data latent in your activity history.
From looking at dozens of customer activity databases, a wonderfully human, complex picture emerges. Guess what? We’re pretty darned diverse in terms of our preferences and email response behaviors. They are not the same as everyone else, and they change over time.
At Motiva AI, we see clumps and clusters of open behavior across customers, both B2B and B2C. Some patterns emerge, but they aren’t necessarily stable. Holidays happen, work gets in the way, life happens. One person’s favorite time to respond to emails my change over time.
What we find and surface often surprised our customers. For one customer, their hands down best send time clusters in terms of open behavior – and note the plural here – were 2 a.m. Sundays, 9 p.m. Fridays, and 7 a.m. Tuesdays in that order. This very large global B2B customer has lots of North American time zones they’re dealing with, but those weekend slots shocked this team.
Another enterprise customer in the gaming industry has an even more fractured set of open preferences. There are literally two dozen or more open preference clusters spread across global time zones.
Despite these varying preferences, the current approach is generally to guess a single golden slot (9 a.m. Tuesday! Business hours M-F only!) across a population, and then to make that guess for each time zone represented in a population and configure your now quite complex email campaign accordingly. Campaign is deployed, and you hope that this is not too far off the mark.
Sheesh.
For global campaigns, you’re literally composing a half dozen or more parallel campaigns that all do the same thing in order to treat each time zone with your guess. The level of needless complexity and sheer management time required strains many teams. And for what? Very few marketers have time to even review whether their guess was a hit or a miss.
The problem is that marketing teams have been measuring open rates and trying to determine optimal send times based on a supposed average of an entire population, because that’s what technology is limited to, and this is what “experts” out there say is the right answer.
This narrow approach is changing with machine learning coming to the fore and with marketing teams tackling their data more actively. With Motiva AI for Oracle Eloqua for example, machine learning allows us to model those send time preferences us on a per-contact basis – and importantly adapt as people’s preferences change over time. In other words: less guessing and a whole lot more of letting actual customer behavior drive the right approach. There’s also some humility involved here: letting go of the notion that we know what best – and better than customers know themselves.
The best approaches to machine learning for send time optimization should take into account any existing send time preferences available from campaign history. But that’s just the start. Assume that you don’t have any data on much or maybe most of your contact population. Add the fact that preferences you do have could shift at any time. Through adaptive experimentation, machine learning (aided by some clever statistical sampling methods) can address a problem impossible to solve manually and help fill in gaps over time.
The result is gradually improved response rates over time as the machine dials in per-contact send time preferences.
This can happen largely without you as a marketer necessarily understanding exactly how it’s happening. The benefit of having machine assistants on the team leaves the human members of the team to focus on more nuanced questions and tasks: refining segmentation, experimenting with content and channel mix, exploring novel marketing strategy, and more.
At a deeper level, what this means is that we’re beginning to treat customers as individual people, albeit in this very modest way. Also, by taking into account those individuals’ needs and wants, we begin to shift marketing to something not just more personal and adaptive, but more effective. And that’s actually pretty exciting!
Machine-driven Send Time Optimization moves marketing towards being customer-oriented and opens the door to bringing the same kind of approach to content, sequencing, omnichannel design, and marketing strategy. It moves us towards adaptive customer relationship building, based not on the convenience of the marketer, but rather on the needs of your customer.
In the end, it’s about listening to what your customer is telling you. No more hearsay, no more myths about the magical golden hour of send time and no more guessing. Just the data driving the next best thing to do for your customer.