A-B Testing

Measure the performance of your different preferences with A/B testing.

A / B Testi (A / B Testing)

Let’s say you have a web site and sell some perfume products on it. Do you think your rate would increase if the button “Purchase” on your web site was turquoise instead of green? If you have such issues on your mind and would like to “try and see”, then it is easy to get the answers to these questions by means of doing an A / B Testing.

How To Do A / B Testing?

To do the A/B testing, first you need to establish the technical infrastructure for the A/B testing on your web site. This infrastructure could be Google Optimize, Google Analytics, Insider, Optimizely or VWO. Once the infrastructure has been established, we need to identify the hypothesis of the A / B testing. For example, you’re A / B hypothesis could be as follows. “My sales would increase if I added a chat module on my web site.” In order to verify this hypothesis, we have to test it. Once the required technical installations have been completed, we create two different versions, where a chat module is used and not used, with the help of the respective testing tools. For example, we make sure that 50% of the visitors on your web site get the service with a chat module, while 50% of the visitors get the service without a chat module. After a 2-week test, the results may be as follows.

Test Clicks Purchasing Purchasing
With Chat Module 1000 23 %2.3
Without Chat Module 1000 17 %1.7

 

According to these results, adding a Chat module has a positive effect on increasing your sales. Therefore, you can increase your sales by activating the chat module for the entire site with peace of mind.

Considerations When Performing A/B Testing

While designing your test, make sure that your traffic is homogeneously distributed. For example, be sure that 50% of your traffic from Google / CPC goes to the version A, while 50% goes to the version B. Check this for all the traffic sources such as Google Ads (AdWords), Facebook, Instagram, organic traffic and e-mail traffic. If there is no statistically significant difference, then there may be no winner of this A / B testing. For example, if 1000 clicks result in 23 purchases and, 1000 clicks result in 21 purchases, then there is no statistically significant difference between them. Do not test to different variables concurrently. For example, do not carry out a test that the version A contains both an affordable price and a red button, and the version B contains both an expensive price and a blue button. If you need to test these two criteria concurrently, then create 4 variations instead of 2. Make a good decision on the solution that you are to use doing the A / B testing. Make your decision once you have well understood the capabilities of the respective A / B testing platforms. Everything you may wonder about may not be a subject of an A / B testing. For example, a question such as “could I increase my conversion rate if I reduce the opening speed of my web site from 9 seconds to 3 seconds?” should not be a subject of an A / B testing. It would be nothing but a waste of time if you want to test again and again such areas well proven to be useful for the users. Keep on mind that you do not have to rediscover America in an A / B testing.

If you think that you could make use of the capabilities we offer for A / B testing, please feel free to contact us and let’s decide together what we can do for you.