Recommender Comparison Example Configuration

Experiments are created, started, and managed on personalization points. For this experiment example, configure two cohorts — one as the control and the other as a variation — using the recommenders that you created and want to test. Each cohort is configured with 50 percent of the defined test population.

1. From the App Launcher, search for and select Personalization Points.
2. Locate and open the Cart Page Recs personalization point.

   

3. From the Related tab, in the Experiment section, click New.
   The New Experiment page opens and shows any inherited personalization point properties that the experiment uses when generating results.
4. Enter an experiment name, and an optional description.
   The experiment API name is auto-generated.

   

   Parameter	Example Setting
   Experiment Name	Test Recs on Cart
   Experiment API Name	Test_Recs_on_Cart
   Description	Test recommendations for different recommender options.
   Personalization Point API Name	Cart_Page_Recs
   Description	Increase revenue at checkout
   Personalization Type	Recommendations
   Content Schema	Recommendations - Goods Product
   Maximum Number of Recommendations	3
   Source	Personalization App
   Authentication Required	Unchecked

5. Click Next
6. Select the primary metric that you want this experiment to test against.

   This metric determines which variation wins across all cohorts defined for the experiment. For this experiment, select the [product order] Revenue from the list.

   

   For information about configuring the Product Order engagement signal, see Example: Product Order. For information about configuring the revenue engagement signal metric for the Product Order engagement signal, see Example Engagement Signal Metrics.

7. (Optional) Select any secondary metrics in the Available window and move them to the Selected window.
   Secondary metrics are tested along with the primary metric.
8. Click Next.
9. Click Nextagain to skip the creation of targeting rules.
   Targeting rules let you select only a portion of the population for the experiment, instead of using the full population. In this experiment, we let the entire population participate, to gather more data.
10. Define the experiment’s control cohort.
    In this experiment, define the rule-based recommender that you configured as the control cohort and keep the population value at 50%.
11. Specify the introduction text that you want to appear with the recommendations to the control cohort.
    In this experiment, we kept the introduction text the same for both experiences.

    

12. Define a second cohort.
    In this experiment, define the objective-based recommender that you configured as the variation cohort and keep the population value at 50%.
13. Specify the introduction text that you want to appear with the recommendations to the variation cohort.
    In this experiment, we kept the introduction text the same for both experiences.

    

14. Save your work.