Product Recommendation System Using Co-Purchase Data Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing product recommendation systems in the B2B environment fail to effectively account for product relationships independent of customer demographics, as they often rely on demographic sensitivity and trends, which are less relevant in this context.
Innovation Solution
A system and method that creates lists of 'purchased-with' products and orders them meaningfully to recommend additional purchasing opportunities based on product relationships, such as brand names, categories, and catalog pages, without considering customer demographics, using a data structure that populates and orders product relationships from purchase order data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If product recommendation systems rely on customer demographics and trends, then recommendations can be personalized for B2C environments, but the system becomes ineffective for B2B environments where demographic sensitivity and trends are less relevant
Solution Approach 1:
The system changes the fundamental parameters used for recommendations from demographic-based metrics to product-relationship-based metrics. Instead of using customer age, income, or preferences, the system uses purchase order data to identify products frequently bought together, creating universally applicable recommendation criteria that work across both B2C and B2B environments
Solution Approach 2:
The system creates a universal recommendation approach that functions effectively across different business environments (B2C and B2B) by using product co-purchase patterns rather than environment-specific demographic data. The same algorithm serves multiple business contexts without requiring demographic sensitivity
2Measurement precision
If the system analyzes detailed purchase order data to identify product relationships, then recommendation accuracy improves, but system complexity and data processing requirements increase
Solution Approach 1:
The system performs self-organization of purchase order data through automated analysis of co-purchase patterns. The data itself reveals product relationships through frequency-based metrics without requiring complex external analysis frameworks, allowing the system to automatically generate recommendation rules from raw purchase data
Solution Approach 2:
The system creates simplified representations of product relationships by copying and analyzing purchase order patterns to generate purchased-with lists. These lists serve as simplified models that capture essential product relationships without requiring the full complexity of the original purchase data to be processed for each recommendation query
3Productivity
If the system creates comprehensive purchased-with lists for all products, then more recommendation opportunities are identified, but the time and computational resources required increase
Solution Approach 1:
The system performs preliminary analysis of purchase orders to pre-compute and store purchased-with lists for all products before they are needed for recommendations. This advance processing creates ready-to-use recommendation data structures that can be quickly queried during actual recommendation events without requiring real-time analysis of raw purchase data
Data Source
AI summary
A dynamic merchandising system creates for each of a plurality of products in a plurality of purchase orders a list of products purchased together. This information is then used to create ordered lists reflecting relationships between various product attributes, e.g., the relationships between different brand names purchased together, different product categories purchased together, different catalog pages of products purchased together, etc. From these ordered relationship lists information may be selected and presented to the customer for the purpose of directing the customer to additional purchasing opportunities.