March 20, 2017
Email marketers spend their days trying to make their email marketing campaigns more effective. We all want our email campaigns to engage our customers and drive them towards our goals, such as conversion, but what’s the best email strategy? What emails should we send? And when should we send them?
Optimail is in a unique place to address these questions because, unlike other email marketing solutions that allow marketers to build and test only a handful of different strategies, Optimail’s AI explores thousands of strategies automatically to find the strategy that works best for individual customers and segments. This permits us as data scientists the opportunity to take a peek at the data and see which strategies tend to work in general — across campaigns — in order to gain new insight into email campaign optimization.
Today, we use this approach to tackle a major question for any email marketing campaign: How frequently should you email your customers?
To answer this question, we took a random sample of 100,000 emails recently sent out on our platform, most of which belong to onboarding (welcome) campaigns and retention campaigns — prime uses for drip email campaigns. For each email, we know lots of things including, most importantly, which customer received it, which campaign it was part of, and whether it was opened. We also calculated how many days it had been since the customer was previously emailed, whether they opened the previous email in the campaign, and whether they opened the current email (our engagement metric of interest).
We know that lots of factors contribute to whether or not someone opens an email, including features of the campaign, the email, the subscriber, and the timing. To control for as many of these as possible, we measured the influence of email timing using a model that looked like this:
Accounting for all these additional factors aside from email timing allowed us to control for their influence and get a better handle on the importance of email timing alone. For example, we know that emails from some campaigns are opened more than others, that people generally disengage with emails over time, that certain subscribers like to open emails while others do not, etc. By accounting for those general effects, we were able to isolate the influence of our main effect of interest: how many days it had been since the customer was last emailed.
When we started this analysis, I thought we would find one “golden rule” of email timing — the optimal frequency at which to send emails.
But our results were much more interesting.
We found that, on the one hand, when customers had opened and read the previous email in the campaign, emailing them again the next day was the most successful strategy. Their likelihood of opening the emails (i.e., engaging with the campaign) declined over time after that.
On the other hand, when the customer had NOT opened and read the previous email in the campaign, the most successful strategy was to wait 4 days and then try again. Emailing too soon led to lower-than-average open rates. By not opening the email, the customer was giving an important signal: now is not a good time.
Importantly, these findings are not just average open rates; they are output of the model above, meaning they are estimates that control for the influence of many extraneous features that influence email-opening.
Broadly speaking, these results once again highlight the importance of tailoring your email marketing campaigns to individual customer’s behaviors and preferences. Treating every customer in the same way (e.g., with the same email timings) means that your marketing campaign will necessarily be ineffective for some people.
This is why we built Optimail to adapt to individual customers and behaviours using artificial intelligence. Why not try it out and see what it can do for you? In the least, stay tuned to the blog for more insights!