Skip to main content
Product recommendation capabilities are a premium feature. Contact your Customer Success Manager to enable them on your Kameleoon account. Access the Product recommendations workspace by navigating to Settings > Recommendations.
A recommendation strategy defines how you select and filter products to display. When you create a strategy, enter a name and description, and define the maximum number of products to display in the block. After you save the strategy, configure its rules, templates, and design.
Click Recommendations > Recommendation strategy. Then, click your recommendation block > Copy code to generate your block’s snippet. Paste it into your page’s source code where you want the block to appear.Alternatively, use Kameleoon’s Activation API to serve product recommendations. The API uses the latest saved template version and returns a JSON object containing the recommended products and the template’s HTML. The API response does not include CSS; you must apply styling to the output manually.When implementing your block’s code, include the data-recommender-code (product recommendation block ID) parameter. For blocks using algorithms like Similar and They also buy this, include the data-recommender-item parameter to specify the current product’s ID.
Customize the frequency and specific timing of your product data feed updates in Store settings.
No. Product recommendations support web and app channels only. You can’t push dynamic recommendation widgets directly to email templates.
Yes, you can preview and simulate your product recommendations on your website using the Widget Studio.To do this, create a widget that contains your product recommendation block, then add it to your experiment. Once configured, use the simulation feature in the Widget Studio to visualize the widget. Simulating allows you to verify both the design and the products the algorithm returns before launching.
Kameleoon calculates normalized_price, normalized_purchase, and normalized_margin as intermediate values it uses to calculate the sales rate. Kameleoon calculates all three using the same normalization formula, applied separately to each product’s price, number of purchases, and margin.To normalize a metric, Kameleoon first calculates the vector length across all products: the square root of the sum of each product’s squared value for that metric. It then divides each product’s raw value by that vector length, which produces a normalized value between 0 and 1.For example, with two products priced at 100,000 and 500,000:
Kameleoon then combines the three normalized values using the same weight coefficients as the sales rate calculation: 1 for normalized_purchase, 0.8 for normalized_margin, and 0.05 for normalized_price.
By default, Kameleoon calculates the sales rate over a 1-month period. Customize this period in Store settings, under Sales rate calculation.
The sales rate is primarily a ranking and sorting signal. Kameleoon uses it in two places:
  • Product recommendations: The Popular algorithm sorts recommended products by sales rate.
  • Search: Pass sort_by=sales_rate in the Search API to sort search results by sales rate.
No. Regular expressions (regex) aren’t available for defining complex matching logic within recommendation rules.
No. The platform doesn’t support Bring Your Own Algorithm (BYOA). Use the suite of built-in algorithms provided by Kameleoon.
While Kameleoon doesn’t enforce a strict limit, requesting many products impacts algorithm performance and increases response times. Display fewer than 50 products per block. To display larger product sets, use Collections.
No, some algorithms only work under specific conditions. For example, the Similar products algorithm, the Products usually purchased with this item algorithm, and the Store recommendations algorithm only work on product detail pages. The Products usually viewed with this product algorithm only works if the cart contains items.
Yes. You can use the product size as an HTML variable in your template, provided the size appears as a parameter in the data feed.
No. The Store Recommendations algorithm specifically requires the accessories field in your data feed. It doesn’t recognize custom field names for manual product relationships.To manually link products (for example, to display complementary items), map the related product IDs to the accessories field.The algorithm displays only the products linked via the accessories tag. You can restrict these results further by applying standard filters to the recommendation block (for example, to exclude out-of-stock items from the manual list).
To use the Currently on wishlist algorithm, explicitly track when a user adds an item to their wishlist. The recommendation engine doesn’t automatically detect these interactions without a specific API signal.Implement the trackAddToWishList() method from the Activation API. This method sends the necessary event data to Kameleoon, which populates the wishlist algorithm.
Supported attributes depend on your feed format. Note the following limitations regarding labels and stock data:Custom labels (custom_label_0 - custom_label_4)
  • Not supported.
  • The recommendation module does not import custom_label attributes from Google Merchant Center feeds. You cannot use these fields for filtering or display.
Stock quantity (stockquantity)
  • Google Merchant Feeds: Not supported. The Google product data specification does not include a numerical stock count (only “availability” status). Therefore, you cannot import exact inventory numbers via Google CSV/XML.
  • Kameleoon XML: Supported. To use exact stock counts, use the native Kameleoon XML format and populate the stockquantity field.
  • Viewing and using stock data: If you import stock data via Kameleoon XML, the values appear in the Data feed overview. Use the stockquantity attribute to create filters (for example, stockquantity > 5).
Base your decision on your catalog size and technical infrastructure. Use the XML import method if your catalog contains fewer than 50,000 products and you lack a website data layer. Conversely, use the Product Recommendation API if your catalog exceeds 50,000 products.