What is A/B testing and how can it make your website better?

One of the best things about websites is that they are adaptable. Through updates to content, designs or functionality your website can be changed to better inform your stakeholders and engage your customers.

However, what does the most optimised version of your website look like? Do you need different images to promote better engagement? If you change the layout of your forms will they actually be easier to use? Will a different colour palette on across your website result in a calmer interface and better conversions? Or, will your proposed changes break everything you have worked so hard to build?

In the sections below we look at what A/B testing is, how you can structure your next test and what kind of things you should be looking to optimise using this approach.

How does A/B testing work?

A/B testing, also called multi-variant testing, aims to present multiple versions of a piece of content to an audience and then monitors the results. Ideally, this is done ‘blind’ where the user does not know they are involved in an experiment and the experimenter has no control over which version the user might see.

In this way, you remove much of the bias from your A/B testing and you are left with actual usage data that can help you understand what variant actually is the best way for your content to be presented.

While this sounds complicated there are many tools and templates available to run your A/B tests and make sure you arrive at the best possible outcomes.

Starting a new experiment with Google Optimize

We recommend using Google Optimize for A/B testing as it works on almost any website and integrates directly with Google Analytics making additional analysis much easier.

To set up an experiment you will need to indicate the page you are looking to optimise and the metric you are aiming to improve. Once you have done this Optimize will create a design interface based on the ‘Original’ page, allowing you to build new design variants directly.

Controlling the traffic

Once you are finished designing you can run the experiment with Google automatically detecting and distributing your website traffic to all your different design variants, and collecting data the whole time. Optimize gives you a lot of control so that you can run the experiment without impacting the normal flow of your website. For example, you can allocate the percentage of traffic that gets sent to the ‘Original’ and other variants or you can limit the experiment only to traffic coming from a specific GoogleAd.

Experiments need to run for 30 – 90 days depending on how many variants you are testing and how much traffic moves across your website.

Analysing your results

Once your experiment is complete Optimize will have a lot of data for you to review. Due to the integration with Analytics, a lot of this will be transferred over there for analysis but Optimize will automatically give you two very important metrics, ‘likely to be better than the original’ and ‘likely to be the best variant’. These are expressed as confidence scores where the higher the number, the more confident you can be that a given variant will outperform the original version or be the best performing variant in the experiment.

Structuring your A/B test

While tools like Google Optimize can assist with running and interpreting an optimisation experiment it is still important to spend time considering what you are trying to optimise and what any existing data might be able to tell you.

The first step is to identify pages that are not performing optimally, for this you will need to turn to analytics tools such as Google Analytics.

A good place to start would be to look at pages on your website where you want users to complete an action, such as:

• Watch a video
• Submit a form
• Complete a sale
• Download a file

Alternatively, you may want to start with a page that you can see is performing poorly due to:

• High bounce rate
• Low dwell time
• High exit percentage (without action)

Then you need to start framing your experiment by proposing a testable hypothesis and identifying a measurement of success. For example, I might hypothesise that I am seeing high levels of cart abandonment because users can’t work out how to move from the cart to the checkout.

In this case, I have identified the problem (cart abandonment), proposed a solution (simplified user experience from cart to checkout) and there is an inherent marker of success, increased sales.

Your proposed solution will likely exist in many forms, in this case we could, for example, make the buttons bigger, a different colour or both. It is now time to start listing all the possible ways you could alter your design, content or layout to arrive at a possible solution.

This could potentially become a large list so we would recommend keeping the list to three or four things at a time. Too many variants will make your experiment take longer and reduce confidence in your results.

You should now be ready to run your experiment!

What parts of my website can I A/B test?

The great thing about websites is just how flexible they can be and how much you really can trial in an A/B testing format. We have included some ideas below but you are really only limited by your imagination and creativity!

Change your layout

Often times, the simpler your layout, the easier your users can find your CTA but you also have to find the balance between simplicity, branding and functionality. Test moving the CTA or your buttons around or adding and removing other content on the page.

Improve your website navigation

If users are getting stuck in a part of your website, maybe you have issues with your menus or navigation options. Test different menus, fonts and layouts – even adding secondary menus to help users navigate your site. Of course, you can go the other way too, removing navigation entirely if you want users to stay focussed on a task such as completing a sale.

Test your offers

Use A/B testing to deliver different offers or different versions of the same offer to see what resonates with your audience. This kind of test can lead to another area of optimisation called Conversion Rate Optimisation, or CRO.

Create skim-able headlines

Users do not read websites, they skim them, so making sure your headings stand out and catch users is vital. A/B testing can allow you to trial lots of different headlines and then settle on the version that sees the best engagement.

Focus on the Call To Action

Your ‘Call To Action’ is the likely the most important element of any page on your website, so getting this right should be a priority. Test different copy, layout, buttons and offers in combination to find the winning variation.

Test appealing images and videos

Will a stylised product shot, a video or a simple side-on photo better engage your audience and convert more sales? In a world of skim-able websites your images need to be engaging your audience to make them slow down and take in your content and products.

Minimise form fields

Optimising forms is a balance between getting the information you need in the fewest fields as possible. Use A/B testing to deliver different versions of your key forms to your audience and see which variants work best.

Optimising your web pages can have an enormous effect. There are case studies all over the internet showing amazing results, such as this 550% increased conversion rate, but it is important to remember to plan carefully and to actually follow through with experiments and learn from the data you generate.

Make A/B testing a habit

We strongly believe every business should be testing and optimising their web pages constantly. Iterative development is one of the best ways to drive continuous improvement on your website and as it is often achieved through content manipulation it can be a cheap and quick way to achieve performance gains. If you can find a way to build A/B testing into your content generation you will start a cycle of iterative improvement through the analysis of variant performance.