Multiple enhancements have been provided for configurable strategies. The Similar Products model has been added to Configurable Strategies. This model finds similar products based upon a comparison of text attributes and descriptions. It can now be used with other options available with Configurable Strategies, including different ways to seed the strategy using the context or user history, sorting or re-ranking the recommendations based upon user affinity, and applying category diversity or category siloing filters to the results.
A new version of the New Arrivals model has been created and added to Configurable Strategies. This New Arrivals model includes all products sorted by recency, where the current New Arrivals model sorts by top selling new arrivals. With the new model, New Arrivals (Recency), you can also then apply different seed, sorting, and filtering options to be more relevant to each user.
If you want to use our deep learning system with Natural Language Processing (NLP) then 2 additional models are also available to you, NLP Similarity and NLP Cross-sell. With these models the deep learning system is capturing the feature vectors for products based upon their text descriptions and trained with viewed together or bought together events, enabling similar products and relevant cross-sell products to be determined even if products are new and have no historical data. You can also personalized based upon affinities to these feature vectors using our Vector Model Service (VMS).
Dynamically re-rank recommendations by affinity to feature vectors using the Vector Model Service:
Additional report visualizations and metrics have been released. For Site Analytics, you can identify the performance for combinations of Channel and Strategy, Page Type and Strategy, Placement and Strategy, Categories and Channel, Categories and Page Types, Categories and Placements, and Categories and Strategies.
In addition, the metrics Items from Recs and Orders from Recs have been added to the Advanced Merchandising report.