Which Is Better AB Tests Or Split URL Tests?
There is no right or wrong answer to this question.
You can read about the difference between AB experiments and split URL experiments here.
A split URL experiment is a specific type of an A/B experiment. In both types, your website visitors (traffic) are split between the control and the variations.
And in both types, FigPii records the number of visitors and conversions each variation generated. The real difference between the two methods is how variations (treatments) are created and served to your website visitors.
Things to consider:
- When you are using AB experiments, you do not have to create the variations on your website. Instead, you will use FigPii to create these variations. So, effectively AB experiments save you time when you are creating your experiments.
- When you are using split URL experiments, you will have to create the variations on your website. Each of the variations will have its own URL on your site. FigPii will split visitors between these variations and record visitors and conversions. In most of the cases, split URL experiments take a little longer to create.
- Because FigPii is used to create the AB experiment, that means FigPii will have to first load the original page to the visitor and then change it to display the variation design. This is called the flicker affect. So, for a few milliseconds, the original page design will flash for website visitors.
Advantages of split URL experiments
- No flicker affect
- Appropriate when the experiment variations are brand new pages on your site.
- Appropriate when the experiment variations are adding new functionality that does not exist on the control
Disadvantages of split URL experiments
- It takes a little longer to implement
- You are required to create brand new pages on your site and host these pages
Advantages of AB experiments
- quick to implement but most AB experiments someone with good knowledge of JS/CSS
- You do not have to create new pages on your site
Disadvantages of AB experiments
- The flickering effect
- Difficult to implement when the experiment variations are adding new functionality that does not exist on the control