UGG.com | A/B Test | Bundle Recommendation by Intent | PDP

Here is an example of how I like to test iteratively with personalization in mind. First, I like to test the existence of a feature. If it already exists, then testing shows the value. If a feature is not something we need to be the same for all users due to code restrictions, then I will test by intent to make sure all of the key audiences (high, medium, low & unknown) respond positively to the feature. My evergreen implementation strategy will only include the audiences with flat or positive uplift.

Hypothesis: If we provide users with an optimized bundle recommendation module on the PDP, then users will be more likely to bundle and revenue will increase.

Primary KPI: Revenue

> Test 1.0: Show/Hide Bundle Recommendation by Intent

This iteration tests the existence of the pairs with recommendation module, and the control won showing the value of this recommendation.

> Test 2.0: Recommendation Algorithms by Intent

This iteration tests the algorithm used to make recommendations, and the variant won.

> Test 3.0: Recommendation Carousel

This iteration tests a single recommendation vs. multiple recommendations in a sliding product carousel. High & Medium intent preferred the control, while low & unknown intent preferred the carousel. Typically, higher intent users know what they are looking for and don't need to explore options like lower intent users.

There is still more that can be tested like single vs. stacked, placement, and new algorithms.

Hope you enjoyed this walkthrough of how I like to iteratively test!