We love when our clients test — and learn — something with their email marketing campaigns. From what works in subject lines to the length of the email to how to best use personalization, testing in email marketing is key to consistently increasing your results.
But there’s an important caveat: All tests are not created equal. In order to really learn from your tests, they need to be both scientifically accurate and statistically significant results. Did that phrase take you back to college? It turns out that statistics class is useful in marketing. Here’s how you do both:
Your test is scientific as long as you’re randomly segmenting your list for these tests and using different segments each time you test something. That means not just sending the first half of the list version A and the second half of the list version B. Some email marketing platforms – like the emfluence Marketing Platform – offer a random segmentation tool. Every time you test, fire up the random segment generator to split up your recipient list into unique segments. This way, no one portion of your list can sway your results (all mothers, all people who start with A, etc.)
Statistical significance means the difference in results for version A and version B are, well, meaningful. A large aspect of making actionable test results is the size of your test segments. Many email marketers use a 10/10/80 split for their campaigns, meaning they’ll send 10% of their list version A, 10% will receive version B. After a day or two, the marketer sends the winning version to the remaining 80% of the list. This is a great way to roll out a known winner to your full email list.
But, 10% of your list may be too small a segment. When you have fewer than 10,000 contacts in your test segment, a handful of outliers can throw off your results and your results can be misleading. That’s the importance of a statistically significant sample size and results variance. (It really is exactly like that class in college.) You can still test with smaller segments, but be sure you use a statistical significance calculator like this one.
Need a little guidance with your own tests? We love testing! Reach out to email@example.com about your own campaigns.