Split Test
Split Test
Definition
A split test (or a redirect test), is an online controlled experiment between two or more versions of the same page using different URLs for each variant. This means you can test a completely different experience (e.g. an innovation test) or replicate similar pages on a different URLs for landing page optimisation.
A/B testing software, such as Google Optimize, randomly distributes traffic to a website between the two different URLs. This minimises the risk of any difference in the conversion rate being due to the nature of the users selected for the experiment and not due to the change in experience.
Slit testing is a common method for conversion rate optimisation because it is a type of randomised controlled experiment. It uses statistical tests to identify the best performing experience based upon the main goal of the page. The performance of each variant is calculated using the same key performance indicators, such as the conversion rate, bounce rate or the number of leads.
1. Advantages of Split Testing:
Split test offer can be a powerful tool for website and landing page optimisation. The main benefits are:
- They are quick to set up as all you need are the URLs for each variant. There is no need to code changes or use a visual editor to make changes to an experience.
- Successful split tests are easy to implement because you can simply redirect to the new URL and then change the URL if necessary.
- They are ideal for landing page optimisation because it’s risk free. If the variant is not successful it doesn’t influence the website’s performance.
- Split tests enable websites to quickly adapt to user preferences and make evidence based decisions on new experiences.
- Provided you have sufficient traffic volumes you can conduct split tests on small cohorts of users.
2. Creating an Hypothesis:
Before creating a split test ensure you have a strong hypothesis based upon our 8 step process for conversion rate optimisation. Avoid relying on gut instinct because around eight out of ten A/B tests fail to beat the default variant. By following a systematic approach to creating split tests you will increase your chances of success. Your hypothesis should clearly explain why the change you are making will influence behaviour and improve your North Star metric.
3. Selecting your Target Audience:
All visitors to the specific page can be included in a split test or rules can be set to limit traffic to predefined segments. Such segments might relate to the type of user (e.g. new users), source of traffic (e.g. referrals), technology (e.g. mobile), location (e.g. country) or a custom segment could be created (e.g. Google Analytics audiences) for the experiment.
Many split tests are restricted to new users to reduce the probability of returning visitors being served the new experience and reacting to the change. This is not full-proof because visitors can clear cookies and some browsers clearly cookies on a regular basis. However, it is good practice to minimise a ‘shock’ response to having seen the default design.
4. Choosing your Metrics for Success:
Each website should have a North Star metric which is closely linked to the goal of the website. This can then be used as the primary KPI for each split test and will help avoid different experiments having conflicting goals. However, you should also choose secondary metrics (e.g. average order value or returns) which provide context and allow you to understand any potential downside to implementing the test result.
5. Running a Split Test:
Google Optimize is a free A/B testing solution which allows you to set up a split test in a matter of minutes. You can implement Optimize solely via Google Tag Manager if you find it difficult to get new coded added to your website. It also allows you to link your Optimize account directly to your Google Analytics view and analyse your test results in either Optimize or Google Analytics.
Optimize now offers you the ability to edit tests without having to stop tests which gives you the ability to correct mistakes during a split test. Simply select ‘Edit’ and then choose ‘Edit running experience’.
6. What to Test?
Split tests are ideal for landing page optimisation because each landing page has a unique URL. This means you can set up landing page tests within a matter of minutes, and because they tend to have high volumes of traffic you can run them in quick succession. Other elements to consider testing with split tests include:
- Headings and taglines
- Value proposition messages
- Images, videos and colour
- Call-to-action buttons and the location of forms
- Acquisition offers and promotions
- PPC landing pages
- Sign up forms
- Pages with a high potential for improvement
7. Split Path Test:
Where a split test continues across multiple pages this is known as a split path test. This allows you to A/B test the whole of an important user journey, such as a sign-up form or checkout process. Below is an example of a split path test conducted on a poker registration form. This kind of test ensures the user experience is consistent throughout the whole journey. Otherwise, making changes on a single page could undermine the effectiveness of the new variant.
Image source:
Resources:
A/B testing tools – Which are the best A/B testing tools?
Customer reviews of A/B testing software – A/B testing tools customer reviews – movers and shakers.
Psychological persuasive techniques – 17 psychological conversion optimisation tips.
Google Optimize – How to set up and run experiments with Google Optimize.
