Category Recommender System Balancing Discovery and Repeat Rankings

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing recommender systems in ecommerce marketplaces are poorly suited for recommending new items and fail to encourage customers to explore new categories, leading to missed opportunities for retailers and lower revenues and customer satisfaction.

Innovation Solution

A category recommender system that uses a combination of discovery category, repeat category, and item recommendation engines, employing collaborative filtering and machine learning models to determine personalized rankings and merge recommendations based on customer segments, balancing repeat and new item presentations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If traditional recommender systems use historical data and traditional algorithms to recommend items, then the system can provide recommendations based on past customer behavior, but the system fails to effectively recommend new items and encourage exploration of new categories

Engineering Contradiction:

Improveability to recommend new items and new categoriesVSAvoidcustomer exploration and discovery

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent segments the recommendation process into multiple specialized engines: a discovery category engine for new categories, a repeat category engine for existing categories, and an item recommendation engine. Each engine handles specific aspects of recommendations, allowing the system to balance between recommending new items and repeat purchases effectively

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system dynamically adjusts recommendation strategies based on customer segment identification. Different customer segments receive different blending ratios of discovery versus repeat categories, allowing the system to adapt recommendations to individual customer behaviors and preferences in real-time

Inventive Principle:

Principle #15Dynamics

2Reliability

If the recommender system focuses on recommending items based on historical data, then the system can maintain consistency with customer preferences, but the system limits awareness of new items and reduces revenue opportunities

Engineering Contradiction:

Improverecommendation accuracy based on historyVSAvoidrevenue and customer satisfaction

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The system changes the parameter of recommendation composition by dynamically adjusting the blend ratio between discovery categories and repeat categories based on customer segment. This allows the system to maintain reliability for each customer type while optimizing overall revenue through controlled exposure to new items

Inventive Principle:

Principle #35Parameter changes

3Productivity

If the system uses simple recommendation algorithms, then the system can operate efficiently, but the system cannot effectively balance repeat and new item recommendations

Engineering Contradiction:

Improvesystem efficiencyVSAvoidrecommendation system structure

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent divides the complex recommendation task into separate functional engines (discovery category engine, repeat category engine, item recommendation engine), where each engine focuses on a specific aspect. This segmentation manages complexity by creating modular, specialized components rather than one monolithic complex system

Inventive Principle:

Principle #1Segmentation

4Ease of operation

If the recommender system does not use customer segmentation, then the system can provide uniform recommendations to all customers, but the system cannot personalize recommendations to maximize exploration and sales

Engineering Contradiction:

Improveuniform recommendation deliveryVSAvoidpersonalized exploration and revenue

Core Design Contradiction:

Ease of operationVSProductivity

Solution Approach 1:

The system dynamically determines customer segments based on transaction history and adjusts the recommendation blend accordingly. Newer customers receive higher proportions of discovery categories while established customers receive more repeat categories, creating personalized experiences that maximize both exploration and revenue for each segment

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS11663645B2Methods and apparatuses for determining personalized recommendations using customer segmentation

Publication Date: 2023.05.30 WALMART APOLLO LLC

AI summary

A category recommender system includes a computing device configured to obtain customer information characterizing a customer's interactions on an ecommerce marketplace. The computing device is further configured to determine discovery category rankings for each category of items available on the ecommerce marketplace for the customer based on the customer information, determine repeat category rankings for each category of items available on the ecommerce marketplace for the customer based on the customer information, and determine item rankings for each item in the ecommerce marketplace based on the customer information. The computing device is further configured to merge the discovery category rankings, the repeat category rankings and the item recommendation rankings into final recommendations based on one or more predetermined merging criteria and provide the final recommendations to the customer.