Beginner’s Guide to A/B Split Testing
Apples or oranges? Mayo or Miracle Whip? Salad or hamburger? Life is full of decisions, which in most cases, get made in an instant without second thought. But, as marketers, our decisions greatly impact open rates, conversions, sales, company branding and reputation. So, which will it be … green logo or red logo? Two-column or single-column email newsletter? Everything matters.
Our own personal feelings or intuition can only take us so far. Just because you prefer the red logo doesn’t always mean your subscribers will react the same way. So, why not ask them? By creating an A/B split test you can learn firsthand how your subscribers will react to your test subject.
A/B testing is simply a test wherein you divide your list into two groups modifying the content for each group to figure out which marketing variables give you better response.
With two versions of a marketing element (A and B), the goal is to find a way to quantify and define success for each. For example, you have two email layouts and you would like to find out which one is better at getting people to sign up to your newsletter. You send one layout to half the people on your email list, and the other layout to the remaining half on your list. Then see how many sign ups you get from each. You then ultimately select the version that works better.
In the end, this saves you a lot of time and money when doing e-mail campaigns because it tells you what works and what does not. It allows you to fine tune your email format, call-to-action buttons and whatever elements you want to test. More than that, it increases both your conversion rates and return-on-investment.
But how do you do this? Here are 10 tips on how to start your own A/B testing campaign:
1. One at a time, please.
When you are starting out, it is easy to identify a lot of variables to test thinking that if you test them all at the same time you could save costs. But this could only lead to confusion on which element really works for each group. For example, did you get more signups in Group A because the layout was better, or was it because of the bright design of the call-to-action button? Test only one variable at a time to be sure.
This is also important if you are going to test email campaigns that lead to the same landing page. You would not know what caused the higher conversion rates if you conduct more than one test at a time.
2. Test the minor revisions.
Some marketers think that they should only conduct A/B testing when they do a complete revamp of their website or create a wholly different layout. But in email marketing, the slightest changes you make to the layout, or font, or colors could affect your conversion rates in a big way.
3. Check out other data too.
To do A/B testing, you should identify a single measure of success. For example, one call-to-action button took more people to your landing page, but the other one gathered more sales from the few people that it took to your landing page. Which one weighs more to you?
In this case, you might realize that the first button might not be better as it may have given your recipients wrong impressions about your products that led them to click on it, and when they got to your landing page, they found that it is not exactly what they are looking for.
4. Split your email list randomly.
The original version of your email format, call-to-action button or landing page would be your control group, while the newer one will become your treatment group.
You should be sending one version to half of your email list and the other version to the second half. Ideally, these two halves should be more or less equal such as how long these recipients have been in the list, demographics such as gender, age and income, and other characteristics.
Admittedly, going through all these is going to take a lot of time. The good news is that you can just randomly assign recipients to either half and still get valid results.
5. Test both groups simultaneously.
Timing affects your conversion rates. The day and time you sent it, or which month of the year you sent it, would affect whether your recipients would open it or not.
If you do not test both of your groups simultaneously, you would never know if the success of one was due to the call-to-action button, or the different time of day they received it.
6. Just how much better is better?
Let’s say that your control group has 75 signups, while your treatment group has 77 signups. Is the treatment group necessarily better than the control group?
Be sure that you brush up on your statistics and check out the statistical significance of the difference in the two groups. Aim for a high statistical significance: 97% – 99% is typical. Again, the Internet comes to the rescue of the numerically challenged marketer: there are online tools that can help you test the significance of your results such as this one: http://visualwebsiteoptimizer.com/ab-split-significance-calculator.
7. Remember your branding.
If you have been using the same format for your emails, then you might want to check whether the changes you want to implement would shock or distract your recipients. This is especially true for your website. If you have a strong and regular following for your website, you might want to test it out on new visitors. This is because the variations might not sit well with your branding and these might not even be implemented in the end.
8. Do not preempt the results with what you expect.
You might be surprised to find out that the winner of the A/B testing you have is the dark horse. For example, you might think that the new call-to-action button is ugly and not easy on the eyes. But you may be surprised to find out that it results to more conversions. Not preempting makes you more open to implementing the changes.
9. Set an appropriate time frame for your A/B testing.
You should take a look at your past email campaigns to see just how long an A/B testing should last. Ending your testing too early might mean that you miss out on the signups that occur after that date, while making it drag on a day or two too long would mean that you are serving your email recipients a format that is not converting too well and you lose out on these conversions.
10. Make it consistent.
You might be testing a call-to-action button on your e-mail that also appears on your landing page and in other places on your website. Make sure that this call-to-action is consistent across all places where your recipients can see it, so that your results will be conclusive.
Most marketers will tell you that A/B testing is difficult to do when you are starting out, and you are bound to make mistakes. But with these tips, you are sure to avoid making the usual mistakes that most marketers commit.