Data Sources

Configuring Merchandizing Algorithm

Data Source Types

Data Sources are categorized into 4 primary types-

1. Similar Product: optimized to display similar product recommendations and is typically used in conjunction with Product Page Widgets.

2. Cross Sell: optimized to display cross sell product recommendations and is typically used in conjunction with Product Page Widgets.

3. Personalized: optimized to display recommendations based on the most recent used activity and is typically used in conjunction with Home Page Widgets.

4. Most Trending: optimized to display recommendations based on the overall engagement on the website and is typically used in conjunction with Home Page Widgets.

Data Source Settings

The configuration of a Data Source varies depending on its type. The following settings are available:

1. Baseline

   • This setting is applicable to both Personalized and Most Trending Data Sources.

   • Personalized Data Source: By default, this source provides recommendations based on the most recent user activity. The Baseline option allows you to select an algorithm for first-time users, ensuring relevant recommendations from the outset.

   • Most Trending Data Source: This source typically displays recommendations based on overall engagement on the website. The Baseline option is useful when there isn't enough data, such as during the initial launch, allowing you to select an algorithm to guide the recommendations.

   • The Baseline feature offers seven different algorithms:

     1. Best Sellers: Shows products with the highest sales.

     2. Most Engaged: Features products with the highest user engagement.

     3. Highest Converted: Displays products with the highest conversion rates.

     4. Highest Inventory: Prioritizes products with the most available stock.

     5. Margin: Focuses on products with the highest profit margin.

     6. Filter: Shows products based on the Filter Conditions.

     7. No Baseline: Empty Carousel

   • You can override the default behavior of these data sources by enabling the Use Baseline Only toggle. Once activated, the Personalized and Most Trending Data Sources will display products based solely on the selected Baseline rule.

   • Example: To feature a specific collection page in a homepage carousel, you can choose the Filter option, select the desired collection page, and enable the Use Baseline Only button. This ensures that the data source exclusively displays products from the chosen collection page.

2. Rules

   • This setting is applicable to both Similar Product and Cross-Sell Data Sources.

   • Use this section to create rules that control the recommended products. The rules are based on If-Else conditional logic, meaning if the specified condition is met, the defined result will be executed.

   • You can create multiple If conditions to refine the decision-making process. These If conditions can be combined using Boolean logic (e.g., "and," "or") to achieve the desired outcomes.

   • Example: If you want to create a condition for women's pants, you would define two If conditions using the gender and category fields, and then combine these conditions with "and" logic to ensure the recommendation is accurate.

     
   • Please be aware that Condition Names are case-sensitive, so ensure they are entered accurately.

   • The outcome of each condition is determined by the Return Condition.

   • Example- To return recommendations for women’s tops, you would set up two Return conditions using the gender and category fields, and then combine these conditions with "and" logic.

     • gender equals women

     • And category equals tops

     
   • Rules can be executed in two methods-

     1. Apply to All

        • This method does not immediately return products. Instead, it evaluates all matching rules and then combines the results using the 'AND' rule.

        • The final recommendations will meet all the return conditions of the executed rules.

     2. Create Carousel

        • This method returns products immediately based on the matching condition.

        • Rules listed at the top of the list have higher priority, meaning that if multiple rules affect the same products, only the first matching condition from the top will be executed.

        • When using the Create Carousel method, you can specify the number of products and their ranking through Count and Ranking, respectively.

        • Example: For a Women's Pants Bundle recommendation where you want to return one top and one jewelry product, you would define multiple return conditions. (check the image above)

        • Ranking Provides you 10 ranking options

          1. Relevancy: Considers relevance based on matching visual features.

          2. Relevancy (No Personalization): Relevance based on visual features without personalized data.

          3. New Arrivals: Displays the latest products.

          4. Best Sellers: Shows products with the highest sales.

          5. Highest Inventory: Prioritizes products with the most available stock.

          6. FBT (Frequently Bought Together): Highlights products commonly purchased together.

          7. Margin: Focuses on products with the highest profit margin.

          8. Most Engaged: Features products with the highest user engagement.

          9. Highest Converted: Displays products with the highest conversion rates.

          10. No Ranking: Randomly selects products from the pool.

   • Return Conditions have two additional Settings-

     1. Apply Filter Rule

        • This setting incorporates conditions specified in the 'Filter' and 'Filter Exclude' sections. It ensures that products are evaluated based on these additional filters when determining the final recommendations.

          
     2. Group Field

        • This setting enables you to group products based on tags.

        • For example, in a Bundle recommendation where you want to select matching tops and bottoms, you can use the Group Field to specify the tag-based grouping. You can also define the count of products in each group, which determines how many products are included from that specific group.

3. Filter and Filter Exclude

   • Use the Filter and Filter Exclude settings to include or exclude specific products from recommendations based on general criteria. For instance, you can configure these settings to exclude products from old seasons, gift cards, or other categories.

   • Must Exist: This setting allows you to establish a fallback strategy for applying filters.

   • If the Must Exist toggle is on, all recommendations must adhere to the specified conditions.

   • If the Must Exist toggle is off, the system will first attempt to fetch products meeting all the conditions. If it fails to retrieve enough products, it will relax the conditions where the Must Exist toggle is off, potentially including products that do not meet all the specified criteria.

4. Weight

   • This section enables you to control the importance of each attribute for recommendations.

   • You can adjust these attributes for baseline recommendations as well as for user events such as Product Click, Filter Click, Add to Cart, and Purchase.

   • Frenzy AI models continuously monitor user activity and automatically adjust these parameters to optimize performance.

   • Please note that the Tokenization setting is not available for all field types.

5. Extra Properties

   This section offers various controls for fine-tuning recommendations:

   • Personalization: Toggle to activate or deactivate personalization features.

   • No Carousel: Toggle to show or hide the carousel display.

   • Display Same Product on PDP Page: Toggle to include the PDP (Product Detail Page) product in the recommendations. This feature is particularly useful for Bundle recommendations, as it displays the PDP product as the first item in the recommendations.

   • Product Pool Size: Adjust the number of products considered for personalization.

     Example: When using the Most Trending Data Source with Filter as the Baseline to display a specific collection page, setting the Product Pool Size to 10 will ensure that personalization is applied only to the first 10 products of that collection.