Implementing Winning Tests

Implementing a winning test template

Once a test is ended, you will have the choice to apply your test variant as your new default experience for your homepage or for the assigned collections, products, or pages. To do this, click Apply variant in the top right corner of your test. Your store will now use your variant template as the default experience for all products, collections, or pages originally assigned to your control.

How winning test templates are implemented

Depending on the type of template you've tested, the implementation method may vary.

Homepage (Index) templates

Unlike other template types on your store, homepage templates are not "assigned" - your store will always use a suffixless index template (i.e. index.liquid or index.json) as your default homepage template.

This means that in order to implement a suffixed homepage template as your new default experience, Shoplift must replace the JSON or liquid code from your original, suffixless template (index.json) with the JSON or liquid code from your tested, suffixed template (index.sl-AE2FB8F1.json).

However, we don't delete the original code found in your older, suffixless index template. Instead, we migrate it to a new template that contains the date of implementation as the suffix, like "index.05-04-2024.json".

If you would like to revert to your original homepage template at any time, simply replace the code in your index.json template with the code from your previous template, now named with the date of implementation as the suffix.

Collection, product, and page templates

When a tested collection, product, or page template is implemented via a winning test, Shoplift simply reassigns whichever collections, products, or pages are currently assigned to your original (control) template to your variant (tested) template. Your original template is unchanged, and can be reassigned at any time.

You can still choose to implement the test variant of an inconclusive test as the new default experience. However, this is not recommended because the performance change between templates is not statistically significant.