Split testing: what is it and who needs it?

Split testing: what is it and who needs it?

Split testing (or A / B testing) is a comparison of two or more page variants in order to choose the most optimal one.

Tests allow you not to act blindly and guess which heading is better, what color the button should be; what happens if prices go up, how does a particular stock work. And they allow you to check and measure how much the changes made to the site page affect the conversion and your profit.

How is split testing done?

To make it clearer, let’s take an example. Let’s say you have two identical versions of the page (A and B), which differ only in the call-to-action button: on page A there is an “Order” button, on page B “Buy”. You alternately show these pages to visitors (moreover, the visitors themselves are not even aware of the existence of the second option), and after a while you analyze on which page the purchases were made more often.

What can you test?
  • Heading
  • Call to action, or call-to-action
  • Selling text
  • Arrangement of elements
  • Prices
  • Forms
  • Special offers and promotions
  • Design
  • Images
  • Video
What testing methods are there?
  • Consistent – first, the results for the first option are collected, then for the second. We do not recommend this testing method because of its huge error.
  • Parallel – both options are tested simultaneously, the tested pages are shown alternately for new visitors. The standard test case that we usually use.
  • Multivariate testing – Simultaneous testing of many different options.
What do you need to know before testing?
  1. Be clear about the purpose of your testing: what we are testing, why we are testing this, why the test should have a positive effect.

Don’t just haphazardly test which red or green button works best. Always formulate the hypothesis you will be testing.

An example of a hypothesis: if you make the title extraordinary, the snippet will attract more attention in the SERP, which in turn will increase traffic to the site.

If you test options without understanding the reasons, then it’s just wasting your time.

  1. Test only one significant element… The more the pages differ, the more difficult it is to understand what determines the behavior of users on the site. Multivariate testing requires professionalism and huge traffic to achieve statistically significant results.
  2. Set a sufficient time span to obtain statistically significant test results. Completing the test too quickly will not be indicative. Depending on the amount of traffic, the test usually takes about 2-5 weeks.
  3. Each test is only true for its specific situation.… For any other hypothesis, it needs to be tested again. If, according to the test results, you realized that a form containing 3 fields works more efficiently than a form containing 5 fields, this does not mean that the same form will be effective for all sites of this subject. For each specific case, a hypothesis must be formed and tested again!
How do you know if the result really matters?

So, at the end of the test, you received the following conversion data:

The page with the Buy button has a higher conversion rate. Does this mean that you need to put such a button on all pages of the site? Let’s check. To calculate the statistical significance, we use the formula:

Y – X> √ (Y + X), where Y> X;

40> 35, which means Y – 40 (Y is always the larger of the indicators).

Substitute in the formula:

40 – 35> √ (40 + 35)

  1. The difference between Y and X is 5
  2. The square root of 40 and 35 is approximately 9
  3. 5 <9, therefore, the equality is not true, and the resulting difference in conversion is not statistically significant.

Be sure to check the results obtained with this formula – this way you will determine whether the obtained data makes sense.

What tools should you use?

There are many tools for conducting an A / B test, both free and paid. For our part, we recommend Google.Analytics. You can start testing in the “Behavior” section by going to the “Experiments” tab.

What are the main benefits of Google Analytics Experiment?
  • Free service
  • Clear, user-friendly and simple interface
  • Automated Statistical Significance – The test simply will not complete until the results are statistically significant.

Split testing is a simple and accurate method for researching a website and landing pages. Using the statistical results obtained thanks to it, you can relatively quickly and easily increase the conversion rate.

Many websites where does not apply split testing, have a conversion rate in the region of 0.3% (but at the same time spend a lot of money on attracting traffic). This is approximately one sale from 400 visitors.

Usage split testing will significantly increase the level of the received conversion. If your conversion rate is 10%, that equates to a 40x increase in visitors, and you get 40 sales instead of 1, without increasing your traffic budget. Working only to improve conversions.

It’s foolish not to use this opportunity, isn’t it?

You can learn more about split testing by getting free video tutorial

PS You can learn about the factors influencing the decision of buyers from the article “How to create the right product card”

Leave a Reply

Your email address will not be published.