Navigational Pattern Recognition for Item Recommendations

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Online marketplaces face challenges in providing personalized item recommendations that accurately reflect users' navigational patterns, often relying on purchasing histories and attributes, which may not capture the full spectrum of user interactions leading to purchases.

Innovation Solution

The system identifies and analyzes navigational patterns from user interactions with online marketplace web pages, comparing them to historical data to recommend items based on statistically significant sequences of interactions, using classifiers to optimize recommendations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If item recommendations are based on purchasing histories and attributes, then recommendations can be provided using available customer data, but the recommendations may not accurately reflect users' navigational patterns and interactions

Engineering Contradiction:

Improveaccuracy of item recommendationsVSAvoidfailure to capture user interactions leading to purchases

Core Design Contradiction:

Measurement precisionVSLoss of information

Solution Approach 1:

The system performs preliminary analysis of navigational patterns by tracking and storing sequences of user interactions (page views, clicks, time spent) before the purchase occurs. This preliminary capture of behavioral data allows the system to later match current user navigation against historical patterns that led to purchases, improving recommendation accuracy without losing valuable interaction information.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements feedback by continuously monitoring user navigational behavior and using it to refine recommendation accuracy. Historical navigational patterns are fed back into the recommendation engine, creating a closed-loop system where user interactions continuously improve the quality of future recommendations beyond what purchasing history alone could provide.

Inventive Principle:

Principle #23Feedback

2Measurement precision

If the system tracks and analyzes detailed navigational patterns, then recommendation accuracy improves, but system complexity increases

Engineering Contradiction:

Improveaccuracy of item recommendationsVSAvoidcomplexity of pattern analysis system

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the complex task of recommendation into distinct components: (1) tracking individual interaction events, (2) sequencing these events into navigational patterns, (3) storing patterns in historical databases, and (4) matching current user behavior against stored patterns. This segmentation manages complexity by breaking down the overall system into manageable, independent modules that can be developed and maintained separately.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system introduces navigational pattern sequences as an intermediary layer between raw user interactions and final recommendations. Rather than directly analyzing individual clicks or page views, the system first aggregates these into meaningful navigational patterns, which then serve as the basis for recommendations. This intermediary structure simplifies the analysis by working with higher-level abstractions.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Adaptability or versatility

If recommendations are based on navigational patterns rather than purchasing history, then user behavior is better captured, but less data is initially available

Engineering Contradiction:

Improvealignment with user behaviorVSAvoidamount of historical data available

Core Design Contradiction:

Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The system performs preliminary data collection by tracking navigational interactions from the moment users arrive at the online marketplace, rather than waiting for purchases to occur. This preliminary tracking ensures that sufficient behavioral data is accumulated before recommendations are generated, enabling pattern-based recommendations even for users with limited purchase history.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system maintains continuous tracking of user navigational behavior across multiple sessions and interactions. Rather than relying on discrete purchase events, the system continuously accumulates interaction data, ensuring a steady stream of useful information for pattern analysis. This continuous data collection compensates for the initially smaller quantity of available data compared to purchase histories.

Inventive Principle:

Principle #20Continuity of useful action

Data Source

PatentUS10846775B1Identifying item recommendations through recognized navigational patterns

Publication Date: 2020.11.24 AMAZON TECH INC

AI summary

Customer interactions with one or more pages associated with an online marketplace may be aggregated into historical navigational patterns, which may be associated with one or more items available for purchase at the online marketplace. Where a series or sequence of customer interactions is subsequently received from a customer, such interactions may also be aggregated into a pattern, which may be compared to one or more of the historical navigational patterns. An item recommendation may be identified for the customer based at least in part on the comparison of the received navigational pattern to one or more of the historical navigational patterns. The item recommendation may include an item previously purchased following one of the historical navigational patterns, or an item related to one or more such items.