You can use Evolv AI to run a traditional A/B test. We'll discuss how and when to use A/B testing.
On this page:
- What is an A/B test?
- Steps to create an A/B test
- Why is the UI different?
- When should I use the A/B test experimentation type?
What is an A/B test?
An A/B test is a type of experimentation where only single ideas are tested against a control. It contains a single experimentation phase and does not automatically change the active ideas.
See how A/B testing compares to Continuous Optimization.
Experimentation Type | Phases | Combinations |
Continuous Optimization | Multi-phase approach: Initialization + Optimization | Automatically creates new combinations of variants to find the top performer. |
A/B Test | Single phase |
Each variant belongs to a single combination for the entirety of the phase. |
Steps to create an A/B test
- Create a new project in the Evolv AI Manager.
- Select A/B Test from the Experimentation Type option.
- Click SAVE and open the project.
- Click MANAGE EXPERIMENTATION and import variables using the project file you saved from the Web Editor.
- Your project file can contain multiple variables. As this is an A/B test, each variant is tested in isolation.
- Your project file can contain multiple variables. As this is an A/B test, each variant is tested in isolation.
- Select an Audience and Optimization Metric, then click Move to Live when ready to deploy.
- Your A/B test is now live.
Why is the UI different?
The first thing you'll notice after launching an A/B Test is a simplified UI.
We only show variant performance instead of combinations because the combinations never change. All ideas are tested in isolation, so the variant performance is all you need.
You still have the overall performance charts to show you how the project version performs. You can use the Variants page to see the individual variant performance.
When should I use the A/B test experimentation type?
You should use the A/B test type if you only want to test single ideas. This includes URL redirects.
- If you want to optimize for the best combination of variants, use Continuous Optimization instead.