- Run experiments that compare different combinations of changes
- Set up a multivariate experiment in Optimizely Classic
- Add sections and variations to the experiment
- Allocate traffic between variations and sections
For an overview of the pros and cons of multivariate experiments, see our article on experiment types.
Multivariate experiments allow you to create a few simple changes, and then test how the different combinations of changes affect performance on the page.
Creating a multivariate experiment
To create a multivariate experiment, start the same way you would to create a regular experiment: load a page into the Editor.
- Once in the editor, click Options > Experiment Type.
- Change the setting from an A/B experiment to a multivariate experiment.
- Next, you'll add variations and sections to your experiment. Each combination of variation and section will be tested in a multivariate experiment.
Click the + Add Variation button to create your variations. In this example, the variations will test the language on the button (the Original, 'Search It!', and 'Google It').
You can change the name of the variation by clicking on the dropdown in the Variation Menu.
After you select Multivariate Test, you will see the + Add Section option above the Variation Menus.
This lets you add a new Section. Sections can be thought of as separate A/B experiments that all run at the same time. Within each section you still have multiple variations.
For the sections, you'll choose the two different images. Click Add Section, then make your changes.
You can change the name of a section by clicking on the name of the section.
Sections are applied from left to right, so the 'Button' section will run first and then the 'Image' section. This is relevant if you happen to be running a multivariate test that affects the same element in both sections.
By clicking on the Variation Menu, then Preview, you can see how the changes will display.
Go to Options > Traffic Allocation and you should see something like this:
What this tells you is that 33.3% of your visitors will see each of the variations from the two sections. They do not have to see the same variation from each section, they will be randomly assigned to one variation from each. You can modify these numbers by just typing in new percentages. Make sure they add up to 100%.
When you go to the results page you will actually see all of the possible combinations that are generated.