Module adjustment and tuning

PS -

 

Adjusting the modules  

It is possible to boost and tune almost every module in the control panel. Boosting and tuning parameters are used to get the desired output and product recommendations. 

 

Examples of how to adjust and tune the algorithms are weighing factors of purchase- and visit history for the individual user or weighting/boosting factors of products that are similar or related to products the user has already seen.  

 

For adjusting the output of the recommendation modules, navigate to Recommendations, hover the Products tabclick Raptor Web Advisor and find the module you want to adjust. You can test the output by putting in the mandatory input parameter in the first box of the module page. You can find the various information directly in the tracking by selecting “Pick Data from live data stream”. Here, you can test with a category/product/brand id. Copy/paste and press “test module”.   

blobid0.png

In the box with the optional input parameters, you will see all the different tuning parameters.  

At the very top of the list, you can include the cookieID in the API call. By including a cookie or userId, the module then becomes personalized to each user. 

When boosting certain products, the weighting factors are used. 

  • VisitHistoryWeightBoosts products that the user has previously clicked on. The value of the weight dictates how much products are boosted relative to the other weights. A high boost of the visithistoryweight would result in the boost of products that the user has previously visited. A high boost in this weight, would then be great in cases where users visit products multiple times before making a purchase. We often see this in fashion, sportswear, or travel.   
  • BuyHistoryWeightboosts products that the user has previously purchased. This weight should be turned up in cases with repeat purchases such as online grocery stores, DIYs stores and in many cases in B2B. 
  • SimilarU2IBoost(look-a-like boost): utilizes twin-analyses to boost products similar to what the user already has seen. Whereas the first two weights to some extent are mutually exclusive, these two weights (SimilarU2IBoostand RelatedU2IWeightare often used together throughout the different industries. 
  • RelatedU2IWeight (Cross-sell boost): used to boost products related to what the user already has purchased. 
  • MerchandisingBoosts: These weights can be used to boost certain products after your choice. This could be products with high margins, products with high stock quantity, products with private label and other use cases. 
  • Serendipity Score: The serendipity score is a factor that adjusts the balance between the popularity and the diversity of products. If certain popular products show up in the recommendations across your website, we recommend increasing the serendipity score to increase diversity in the recommendations across the website 
Have more questions? Submit a request

Comments

Powered by Zendesk