What is an A/B test?

An A/B test, also known as a split test, is a method that compares two versions of content to see which one attracts more visitors. The same piece of content is divided into two versions: a control version A and a test version B. The control version can be an existing piece of content – such as a page or blog post – while the test version contains the same content but with differences in layout or composition. The goal of the test is to find out which version delivers the best performance.

In essence, an A/B test is an experiment in which you present content to users at random. The data from this exercise will often reveal which variant performs better in relation to your content objective.

How does an A/B test work?

Every A/B test can be carried out according to a fixed procedure. You may also be able to use software depending on factors like scale and frequency. Below is a comprehensive, step-by-step plan to set up a successful A/B test:

• Define your goal and hypothesis. Start from a clear objective that reflects exactly what you want to improve about your digital performance. This could be the number of impressions, visitors, clicks, and so on. Next, formulate a concrete hypothesis that outlines which actions are needed to achieve your objective, using KPIs. For example: “the number of clicks on the blue CTA will be 10% higher than on the green CTA.”
• Choose your content. This can be an article or landing page, but also an e-mail or call-to-action. It is worth taking into account the specific characteristics of the content you picked and its objectives. Bear in mind the golden rule that applies to all A/B tests: when choosing your content, test only one element at a time. This will give you a clearer picture of which content element produced a significant difference in performance.
• Create a variant or B version. Using your hypothesis and objective, create a variant or B version of the content item. This could mean adjusting the layout of a blog post, changing the colour or position of a CTA, or modifying certain images – always bearing the golden rule in mind.
• Run the two versions simultaneously. Show both variations to an audience. Split the target audience into two groups: show control version A to one group and variant version B to the other. You may want to use split-testing software, which various marketing automation packages now offer, to place version A alongside version B.
• Collect data from the test results. This can be information about click behaviour, form completions, or time spent on a page. It is also important to track the results using KPIs – pre-defined metrics that translate test results into concrete figures and reflect the performance of the test against the evaluation criteria. Let the test run long enough to gather meaningful insights. TIP: use dashboards and digital tools that give you a clear overview of the relevant KPIs and data to avoid confusing results.
• Compare the statistics results and choose the best-performing version.

What are the different types of A/B tests I can run?

Numerous formats and approaches to run A/B tests exist across a wide range of channels and platforms, including social media, e-mail and websites. The most common A/B tests in content marketing are:

• Titles
• Images
• Content length
• Fill-in forms
• CTAs

1. A/B testing titles  

Whether featured above an article or YouTube video, titles generally have the greatest direct impact on click-through rates and largely determine whether content gets clicked on at all. Is the purpose of an A/B test to find the perfect title? No. But it can give you insight into which title structure, length, and keywords resonate best with your target audience. Select a range of strong options, narrow the shortlist down to the two best and put them to the test.

Also think about your distribution when doing this. Test the titles via e-mail if your site generates most of its visitors through your newsletter. Send one half of your database a message with title A and the other half with title B.

The same goes for social media. Post a message at different times with different titles across different channels to determine the post with the most impact. You can also try this through paid advertising, for example, with sponsored content in two versions: half with title A, the other half with title B. The same applies to social media: post at different times with different titles and track which one performs best. You can also test via paid promotion, for example with two variations of sponsored content.

2. A/B testing images  

We’ve known for some time now that images create more engagement. Images trigger emotion and involvement, and can significantly increase or decrease users’ engagement level with content for that reason. A/B tests can also help you better choose the right images. We are not suggesting you split-test images for every article, but you can find out which types of images generally work best.

For a core topic, you could select a range of images that all relate to your subject from a different angle. For example, you could pick images that consistently feature people, animated or illustrated images, or opt for images that visualise a keyword.

Always ensure that the images are of equal quality, so that this factor does not influence the test’s outcome. Once your image set is ready, you can get started and distribute the same article with different images to compare performance.

3. A/B testing content length  

The ideal content length depends on your target audience, subject, and medium, and can only be reliably determined through structured testing. Does your audience prefer long articles over short ones? Quick videos over full documentaries? Data on how long viewers continue watching your content can already tell you something, but reading time, for example, can be considerably harder to gauge. Tracking time spent by readers on a page via Google Analytics is not always an indication of actual reading.

Some studies – including this one based on Growthbar data – put the ideal article length at around 1,928 words. But you can equally find studies suggesting that long-form content (longreads) of more than 3,000 words is on the rise and that Google’s algorithm likes such pieces.

You could also opt for a more complex set-up and distinguish between your different buyer personas to see how they respond to the content versions. What this looks like can vary considerably depending on factors like subject matter and sector. This is why testing can come in very handy. Address the same topic in 2,000, 4,000 and 8,000 characters. Send these in a split to your database and review the results (not an A/B comparison, but an A/B/C one). You can do the same for short and long videos.nt across segments of your audience.

4. A/B testing forms  

The length, content, and placement of a form have a direct impact on conversion rates and lead generation.

You want to collect leads and perhaps your content is brilliant at that, but the sticking point is your fill-in forms. Luckily, these can be tested too. You can vary the length (the number of fields to fill in), but also the type of questions (which ones are excessive, and which ones should delve deeper instead). Or: where do you place the form on a page?

Next, select a piece of content with a solid conversion rate and send it to your audience with the different forms.

5. A/B testing CTAs  

You can equally be very precise in your split testing for call-to-actions. The wording, design, and positioning of a call-to-action largely determine whether users take action. Whether a CTA lands well is also connected to the associated content. You can test coherence (do the content asset and the CTA actually belong together?), but equally whether a topic is strong enough to come with a CTA that subsequently links to a form.

Beyond testing the actual content, you can equally compare the design of your CTAs or their placement using a split test.