A/B Testing is a specific marketing technique that allows specialists not only to analyze the entire work of a website or an online application, but also to evaluate the effectiveness of the implementation of certain new functions or changes in the areas that are visible to users.
The use of A/B testing is a common practice among marketing experts, business analysts and other specialists that are involved in the project development process.
As usual, this testing performance is required in the common cases, as follows:
the necessity of making some structural and design changes that are essential for the entire operational work of the web source.
the necessity of making some improvements on the web source - there is a list of criteria that are suitable for detailed analysis and further improvement
How A/B testing works
At its bottom, A/B testing is quite simple. There is a typical sequential algorithm of how it works in practice:
Step 1. There is a final selection of components for their implementation or change on the website. In order to make a proper selection of the components, it is better to use one of the convenient techniques for prioritizing testing tasks, such as ICE, RICE, PIE and other prioritization methods.
Step 2. Presentation of the website in the form of its two versions - one with the changes, the other is an old version. The first version, version A (or the ‘control version’), does not contain any changes, but it performs an important function: version A helps specialists receive the most accurate assessment of the effectiveness of those changes that were made in version B.
Step 3. Dividing the website’s audience into two parts, one of which will use version A, and the other, in its turn, will evaluate the changes in version B. It is one of the most difficult part in A/B testing, because there are usually many factors that can have an impact on the final results. For example, groups of audience can interinfluence and use both versions at the same time, but it is a kind of mistake in A/B testing, but it can be fixed by special analysis filters. Also, there are some external factors like times of day or seasonality, so it is necessary to take measurement coincidently at the same point of time.
Step 4. Obtaining and analyzing the testing results. Here it is important to consider two main points. The first thing is that it is pointless to hurry up and wait for the results at a fast clip. A proper A/B testing requires a time to be done and analyzed gradually. Secondly, a singular A/B testing is not enough for absolutely accurate results - it is better to hold one more testing after a while in order to compare two results. It is true that 80% of all primary A/B testings are not statistically significant, so it is important to make more testings for really correct results.
Targets of research in A/B testing
A/B testing can be initiated for different purposes practically in every industry. If it is considered in the context of online projects, there are three typical markers that become the main objectives of this testing.
Marker 1 - Conversion
Here the term ‘conversion’ has its direct meaning: it is the number, which is the ratio of the total number of visitors to the website to the number of active users of the resource who performed a certain action (reading the articles, surfing the website pages, ordering, etc.). Conversion can be presented both as a quantitative result or a percentage.
Marker 2. Economical metrics
Such markers are usually common for e-commerce sites. There A/B testing usually checks such indicators as an average revenue, sales check, etc.
Marker 3. Visitors behavior factors
These metrics include scroll reach (ratio of the number of viewed pages to the total number of the website visitors), session average duration (the average time of the website visiting), bounce rate (the percentage of the users that left the website immediately after its opening), the number of new-time users and other indicators.
A/B testing results
Here is a typical picture of A/B testing on websites:
It is a very simple example of the A/B checking of the total number of the website. After the receiving of data and graphing it all, it is possible to see a certain intermediate result of A/B testing. After that, it is better to continue the testing and allowing both version to work at their personal times.
A/B testing contains many related studies and practices, so if you are interested in this theme, we would like to recommend you to read the following articles: