A/B testing is the procedure of finding out which design works and which doesn't.
AB testing is also known as bucket testing or split testing.
Following are the steps of A/B testing:
In this form of testing, you create two or more versions of your webpage with whatever changes you want to test for.
Both forms of pages are shown to a set of users. You can choose how you want to divide the number of people seeing different pages.
According to the conversion rate and other metrics, a decision is made on which page is doing better.
Finally, the bad performing page is replaced by the better performing page.
This is the main guideline through which the A/B testing works. You can leave the article right there if that is what you wanted to know.
After this, we are going to discuss various things related to A/B testing in detail.
Before diving into how you should do A/B testing let’s first discuss on why should someone care about A/B testing and how is it important for your business.
Say, you are owning a big firm and you change the design of your website and suddenly you start getting 2x number of sales.
You came to know about changing just the design of the website can move the number so radically.
Now let’s say earlier you were making 1000 dollars every day. With the incoming design now you are making 2000 dollars each day.
Now that you know that design is important, you asked your team to make a new design with a few improvements that you thought might be useful for the customer.
The design team was quick and was ready with the solution in 2 weeks.
You shipped the product in the excitement in the morning and went out to a meeting. When you came back at night and looked into the numbers of sales, they went from 2000 dollars to 200 dollars.
Hopefully, you got the point which I was trying to make.
In business, never do something if you think something is going to work. Back your hypothesis with some numbers.
That's the point where A/B testing can help your business.
Rather than having a drastic change in the user experience A/B testing believe in doing it in small iterations. Every time you keep on moving the changes, everything starts showing up to a limited set of users.
Over time the population coverage for all users become 100% which was the initial target.
I have seen some companies keeping separate designs for users depending upon the click history and mouse movement history of the user. Although this is drastic and should not be followed blindly this shows how far are companies willing to go for getting more sales.
With steady changes in the designs of your pages, you can save users from confusions.
It saves companies from losing a tone of money if something goes wrong with the design. Some designs might be offensive for some ethnic background. You will get to know about these things early if your designs changes are only visible to a fewer set of customers.
A/B testing can be as small as changing the header for some users or changing the image for a different set of customer. Size of change doesn’t matter for a change to be an A/B testing.
The existing version of the page is known as the control and the page with the change is known as variation.
In this part, two or more designs are prepared. This can involve any amount of changes in both the version of the same web page. Some companies reveal the new feature to a smaller set of the population (Commonly known as a Beta release of a feature).
For example, WhatsApp has given access to WhatsApp business APIs to very few businesses.
This is done to limit the usage and find the bugs in the platform at an early stage.
Division of the population can be done on any basis. Some try to do it by ethnic background. Some try to do it using geographic data. For some companies, it is random.
Facebook has open sourced this awesome A/B testing tool which can be used to make plans related to A/B testing:
Once done, you can ship the changes to the target population and wait for the numbers to pour in.
There can be any number of metrics that you want to choose for analyzing the usage by the customer. The usage by the customer will define the result of the testing. Was it a failure or a success.
Conversion should be the main factor in deciding the fate of the variation design changes.
You should not keep different versions of the same webpage for too long if they are not meant to be different for different people( We will discuss later). As soon as you find the analyses results either keep the changes or replace the changes.
That’s how the A/B testing framework works.
A/B testing is also very helpful for SEO. Google recognises and rewards your effort to provide better solutions for your customers.
Google itself has made it clear that websites should make use of such feature for improving the user experience. But they always had spoken against clocking.
Clocking is the process in which you divide population on basis of something and show different content to crawlers and different to users.
This can be done by dividing the target population by IP. Google crawlers has a special set of IPs and some websites try to fool crawlers by showing them something which is not there.
People do something like this to hide illegal stuff on there website from the crawlers.
These type of activities will allow search engine to permanently penalize your website.
There are a lot of tools which you can use to make it happen on your blog.
WordPress provides this plugin to help you with the task
If your site is on Jekyll, use this
How can this be beneficial for different type of websites?
E-commerce websites can show different pages to see the difference in the sales for same pages having different designs.
Sports website can choose from various ways of showing the scorecards and player statistics.
In the end, we can say that it is a good practice to make use of A/B testing for your website or business. We recommend following various SEO practices while making use of this awesome technique.
If you liked the post, please share it with your friends and let us know your experiences in the comment section below.