How to conduct A / B testing of website interface?

Evgeniy Krasnitskiy


Web developer

Experience in programming for over 7 years. But even with so much experience, always trying to learn and keep abreast of new trends, methods in development.

How to conduct A / B testing of website interface?

Internet – very mobile sphere, which trends and preferences are changing before our eyes. To commerce site could meet the needs of users and more or less stable to convert them into customers, it is necessary to continuously improve. Change assortment, design, text conversion form – all this and much more able to increase both sales and drop an existing bar. It is to ensure that any site changes inevitably improved conversion and other key performance indicators, and was invented by A / B testing (split testing).
The main advantage of split-testing – the possibility of the most detailed test absolutely every element of the site from the font of the text and ending with a call to action button. After analyzing the data after the test, the marketer can determine what prevents visitors leave the application or perform any other desired action. These “conversion” problems include failure of the button layout, poor text illogical arrangement of blocks of information, too much contact form, irritating background image or something else.
Traditionally, the A / B testing is a fairly uncomplicated scheme: creating a second version of the test page, which are made necessary changes (such as changing a button call to action), targeted traffic is divided in half and poured on both versions. Then comes the time analytics. If the conversion of modified pages increases as compared with the original, so the problem was in the button. If the conversion sags or standing still, then you need to test another interface element. So, gradually testing various elements can be significantly improved key indicators site, making only the “right” change.
Nevertheless, even in split-testing has its own nuances, not taking into account that can be fatal to make a mistake in the results. How correctly to conduct A / B testing interface?
– Make sure that the testing involved a sufficient number of users. Talk about changing conversion makes sense only if through your test page passed all the basic “subspecies” of the target audience. It may take weeks, please be patient.
– Separate traffic sources. The user who went to the site of the content and the user, who went from the mailing list, completely differently motivated. If a regular customer, who has come from a mailing ready to “forgive” you are too detailed conversion form, the “newcomer” out of context, most likely did not make the target action.
– Have a clear idea of what exactly the element you are testing. By changing several parameters at once, you risk getting false results and confused in research. And no need to carry out “test for the test” – you must have a clear understanding of the work and features of the interface to be able to predict any changes in the key indicators.
If you are still trying to improve the site at random and based on conjecture, introduce any innovation and change – your site will not survive in the competition. Do not ignore a powerful web analytics tool – split-testing interface. That he is able to point out the current problems, suggest solutions and to predict the result of any changes.