In a Split-Url test, you can create multiple versions of your website hosted on different URLs. On these versions, traffic is distributed randomly to conduct a randomized controlled experiment. The performance of each of the variants you create in a Split-URL test is monitored and evaluated to identify the most effective variation. A split test enables you to determine which of multiple variations of your web pages—such as landing pages or homepages—has a higher conversion rate for your visitors.
Why split test?
Split testing allows you to evaluate the effects of different website designs, graphics, copy, promotions, and calls to action on user behaviour. For instance, you may compare your current landing page (the control) to one that has been completely changed. After that, you can study the following behaviour of your users:
- Bounce rate – Are consumers visiting the site for longer or are they leaving it sooner?
- Session duration – Do they browse more of the website’s pages?
- Signups/Leads – Are they more likely to perform a conversion activity you’re interested in, such as filling out a form, making a purchase, signing up for a trial, or starting a trial?
Split test vs A/B test
Both Split testing and A/B testing are sometimes used interchangeably. However, what differentiates them is the magnitude of a change one intends to test. Split testing compares a control version of your website to a completely different version hosted on a different URL to see which version users respond to the best. While in an A/B test a control version is compared to a variety of small, single changes of that control version to determine which version performs the best.
Split test vs Multivariate test
In a Multivariate Test(MVT), multiple components of your website are changed at once. The difference between variations and control isn’t large. However, in a Split-URL test, major modifications are made to a variation and evaluated on several URLs.
When should you perform a Split, A/B test, and Multivariate test?
Split tests, A/B tests, and Multivariate tests can work complementary with each other to maximize the effectiveness of the pages on your website.
Split testing is advised when substantial design or backend modifications are made to the website. This is because testing a significantly altered version of your website can demand fundamental adjustments that your marketing may be unable to implement. While A/B testing incorporates changes that are close to the original like changes in headlines, colours, etc., MVT constitutes changes that are even nuanced and are in large numbers on a single page. The magnitude of changes involved in the tests is of the following order –
Split Test > A/B test > Multivariate Test
However, the order of time of conclusion of the tests is in reverse order. The sample size required to detect changes in MVT is a magnitude higher than what is needed in a split test.
So overall split testing works best for implementing significant modifications. Once that significant modification has been performed, further optimizing an existing page may be done using an A/B test and even further with MVT.
What can you split test?
In general, split tests are suitable in case of major changes being made on a website. For instance –
- Backend algorithms
- Business names
- Logo designs
- Website layouts
- Mobile app user interfaces (UI)
- Mobile app icons
- Book cover designs
- E-commerce product descriptions/attributes
- Social media posts
The following are some split test case studies of tests run on VWO:
- Backend technology – Evaluation if HTML5-based website brings more revenue compared to mobile version.
- Increase subscribers – Evaluation of CTA messaging
- Changing product overview page to increase purchases.
When not to use Split-URL Testing?
In a Split Test as visitors land on the control page, VWO randomly chooses either to keep them on the original URL page or redirect them to the variation URL. Avoid a split-URL test while making little modifications to the website, such as changing the colour of the CTA, and headlines as this can cause lag in the variation owing to visitor redirection.
