Split testing
In Sitecore Discover, a split test refers to an experiment where two or more variations of a widget are tested against each other. You run different variations to evaluate which one makes the maximum impact and drive business metrics.
Split testing eliminates the guesswork in website optimization and enables merchandisers to take data-driven decisions.
The following lists common examples of search, recommendation, and personalization use cases for split testing from Discover merchandisers:
-
Does my curated search result for keyword X perform better or worse than Discover AI/ML-driven results?
-
Does burying on-sale products in the search results generate more revenue?
-
Does boosting higher-margin products generate more revenue?
-
What key performance indicator (KPI) delivers the most revenue overall or for keyword X?
-
Does the use of the product attribute short description to determine search relevance generate a higher click-through rate overall or for keyword X?
-
Does recommendation recipe A generate more revenue than recipe B on my homepage?
-
Does the display of on-sale products only on product details pages for products that are on-sale generate more revenue per visit?
-
Does the display of complementary products generate more revenue than similar products in product detail pages?
-
What banner generates a higher click-through rate?
-
What call-to-action generates the highest click-through rate?
Example: Testing recipes against each other
You can create and use the following widget variations of a recommendation widget:
-
Variation with a recipe to display similar items.
-
Variation with a recipe to display trending items.
Example: Testing a merchandising strategy
You can create and use the following widget variations of a search results widget:
-
Variation with a rule to pin bestseller product in the first slot.
-
Variation without any rules.