Any business can benefit from A/B testing to increase revenue and engagement through email. An A/B test can be achieved by sending two different versions of the same email. The study analyzes how changes in the name, content, or subject line can significantly affect your results.
Campaigns that are not A/B-tested have much lower open and click rates. Additionally, they increase revenue.
There are, however, differences between A/B tests. It is crucial to consider how long a test lasts and how it selects a winner to determine its effectiveness.
Consider the conversion you’re attempting
A/B tests are set up by first deciding the campaign goal, and then the intended results. These three things can give you a good idea of how to pick a winning metric for your goals regardless of the number of reasons to choose one metric over the other:
- Your website should receive traffic. If you have an online business or blog, you might host ads on it to generate revenue. You will win a contest if you focus on clicks.
- You should send emails to your subscribers. Sending a newsletter that includes ads that are based on impressions would be our suggestion if you’re just disseminating information. The winning email is determined by how many opens it receives.
- A connected store allows you to sell stuff. When you use email marketing to promote your best-selling products or when you test different incentives to encourage shopper conversions. Revenue should be your winning metric.
Why does this matter? According to our research, you should wait until you are confident in each testing metric before making any decisions.
|Winning Metric||Wait time for 80% accuracy|
The optimal times for each metric are quite different, so don’t waste your time or pick your winner too early! To better understand how we derived our estimates of wait time, let’s examine the data more closely. Here are some reasons why it is so important to use the right winning metrics.
Revenue does not equal clicks and opens
It takes longer to test for revenue than to test for opens or clicks, so you might want to test for revenue instead.
Revenue cannot be predicted more accurately than a coin flip based on opens and clicks!
You are just as likely to choose the winner based on clicks as you are to choose the winner based on the number of clicks. Similarly, open rates can be used to predict revenue. Test if your goal is to generate revenue.
What is the appropriate waiting time?
Almost 450,000 A/B tests from our users with at least 4,000 subscribers were analyzed to determine each winning metric (clicks, opens, and revenue). Comparisons were made between the test winner at the time of the snapshot and the winner at the time of the snapshot.
For each snapshot, we calculated the percentage of predictions that were correct. Wait times of 3 hours and 11+ hours correctly predicted the winner of the all-time title over 80% of the time.
Just one-hour wait times picked the winner 80% of the time, while three-hour wait times picked the winner over 90% of the time. It is still possible to have a winning metric based on clicks even following openings.
Revenue may take the longest to find a winner, but it is perhaps not surprising. Obviously, openings take place first. Clicks from a few of those opens will lead to sales for others people. But, it pays to be patient. You’ll have to wait for nearly 12 hours to correctly choose the winning campaign 80% of the time. A full day of testing is best for 95 % precision.
What important decisions can we make using this data? The following steps should be followed when conducting A/B tests:
- Select the winner based on the metric you want to achieve.
- Clicks and opens aren’t substitutes for revenue.
- Take your time. You’ll be able to make the right choice if you let your tests run long enough. Depending on the type of campaign, you should wait at least 2-3 hours to determine a winner based on opens, 1-2 hours for clicks, and at least 11-12 hours for revenue.
This data is a great starting point, but it is based on several hundred users and may not reflect your own experience.
Test different metrics and durations to determine which yields the best (and most accurate) results for each list using A/B testing.
You might not be able to support our recommended at least 4000 subscribers per combination if your list or segment doesn’t support it. If you would like to formulate future campaign content decisions based upon your entire list, you can do so also.