Refinement Based Personalization API, based on machine learning (ML) algorithm, uses various ML models to capture shopper's data over a period of time and biases the future search results based on the refinements selected by shoppers over a period of time that are established based on their current engagement.
The ML models retain the search query, applied refinements, and the results returned for shoppers with similar searching or browsing patterns. For any future search requests, the ML model biases the search response by utilizing the saved refinements from the shopper's previous visits.
The ML models for Refinement Based Personalization works best for shoppers who are logged in to your e-commerce website. These models need to collect data over the span of for at least 2 weeks to provide personalized suggestions to a shopper. The API will bias the search result based on a selected refinement and other similar refinements if a shopper has established high affinity with a refinement from the rest of the customer data.
Sally searches for "black t-shirt" and applies the 'Size M' refinement. Another shopper Lucy also searches for "black t-shirt" will get a weak bias for 'Size M' in her search results. However, If Lucy applies the 'Size L' refinement, her results will show stronger bias for 'Size L'.
On a returning visit, if Sally searches for a "dress", the results will show stronger bias for 'Size M'. Whereas for a similar query, Lucy's search results will show stronger bias for 'Size L'.
These refinements will be retained by the ML models and be applied to shoppers whose searching or browsing behavior was similar to that of Sally and Lucy.
If a shopper searches for "shoes" and applies the ‘Brand’ refinement as ‘Nike’, the refinements will bias results by ‘Brand Nike’ and other similar brands or categories if a shopper has shown higher affinity towards it from the rest of the customers data.