Personalized Product Sorting via User Affinity Profiles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Users face difficulties in identifying relevant products among numerous options on online retailers' websites, and retailers struggle to provide personalized product information effectively.
Innovation Solution
A personalized category-based product sort system that analyzes users' prior activities and affinities to rank and sort products within specific categories, using weighted affinities and selection strategies to display the most relevant products first, while incorporating diversity and non-personalized strategies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If products are sorted using traditional methods (e.g., alphabetical, price-based), then the sorting process is simple and fast, but the relevance to individual user interests is low
Solution Approach 1:
The system performs preliminary actions by collecting and analyzing user activity data (views, purchases, searches) before the actual product sorting occurs. This pre-analysis creates user affinity profiles that enable personalized sorting without adding complexity to the real-time sorting operation itself.
Solution Approach 2:
The patent introduces an intermediary component (the affinity determination system) that mediates between raw user data and the final product sorting. This intermediary processes user activities and translates them into affinity scores, which then guide the sorting algorithm without requiring direct complex interactions between all user data points and product attributes.
2Ease of operation
If the system displays only personalized products based on user affinity, then user satisfaction increases, but product diversity decreases
Solution Approach 1:
The system applies local quality by differentiating the treatment of different product positions in the display. Highly affinity-matched products are placed in prominent positions (top of list), while other products are distributed throughout the results. This creates varying qualities of presentation across different products based on their relevance to the user.
Solution Approach 2:
The patent implements partial personalization rather than complete personalization. While the system prioritizes personalized products, it deliberately includes non-personalized or diverse products in the results set, applying personalization to only the most relevant portion of the product catalog while maintaining overall diversity.
3Measurement precision
If the system analyzes extensive user activity data to determine affinities, then personalization accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The system extracts only the essential and most relevant features from extensive user activity data (such as purchase history, view patterns, and search queries) rather than processing all available data. This extraction focuses computational resources on the most impactful signals for determining user affinity.
Solution Approach 2:
The patent employs parameter changes by dynamically adjusting the weight and influence of different user activity parameters based on their relevance. Not all user activities are treated equally; the system modifies parameter importance (e.g., giving higher weight to purchases than to casual views) to optimize accuracy while controlling processing requirements.
Data Source
AI summary
Techniques are described for determining personalized category-based product information for a user. The described techniques may in some situations be used by or on behalf of an online retailer to determine a sorted order of at least some products within a product category for the online retailer to display to a customer user on a category-specific Web page of the online retailer, such as based on a browse request by the customer for that category or on another identification of the category by the customer. The determination of particular personalized category-based sorted products for such a customer may in some situations be performed based at least in part on prior activities of the customer, including prior interactions by the customer with the online retailer (e.g., actions by the customer to view and/or purchase particular products, product brands, product sub-categories, product sizes, etc.).