Product Feature Tagging for Real-Time Upsell Recommendations

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Traditional upsell recommendation engines in online retail lack the ability to consider consumers' specific interests in product features, leading to irrelevant recommendations based on stale historical data.

Innovation Solution

Collecting consumer-product interaction data by monitoring the duration consumers view specific product features, and using this data to generate upsell recommendations with similar features-of-interest, by tagging regions of product visual representations and analyzing interaction times.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If traditional recommendation engines use historical purchase and browsing data, then they can generate upsell recommendations, but the recommendations do not consider specific product features of interest to the consumer

Engineering Contradiction:

Improverecommendation relevance to consumer interestsVSAvoidproduct feature interest data

Core Design Contradiction:

Adaptability or versatilityVSLoss of information

Solution Approach 1:

The system performs preliminary tagging of product features and regions before consumer interaction. Visual representations of products are pre-segmented into tagged regions with feature metadata prepared in advance, enabling rapid analysis when consumers interact with specific areas of interest

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent replaces traditional mechanical data collection methods with optical and computational approaches. Instead of relying solely on explicit consumer inputs, the system uses image processing, region tagging, and automated feature recognition to capture and analyze consumer product feature interests

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Adaptability or versatility

If the system monitors consumer interaction with product features in real-time, then it can generate more personalized recommendations, but the system complexity increases

Engineering Contradiction:

Improvepersonalization capabilityVSAvoidmonitoring system complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system segments product visual representations into multiple tagged regions, each representing a distinct feature or component. This segmentation allows the monitoring system to track consumer attention to specific features independently, enabling personalized recommendations without requiring a monolithic complex monitoring system

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces tagged regions as intermediaries between the consumer interaction and the recommendation engine. These regions serve as mediators that structure and organize raw interaction data, making it easier for the recommendation system to process and act on consumer feature interests without direct complex processing of raw mouse movements or view durations

Inventive Principle:

Principle #24Intermediary (Mediator)

3Productivity

If the system uses historical data for recommendations, then it can generate upsell suggestions, but the data becomes stale and less relevant to current consumer interests

Engineering Contradiction:

Improverecommendation generation efficiencyVSAvoiddata freshness

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The system implements continuous data collection and processing, where consumer interaction data is monitored in real-time as consumers browse products. This continuous action ensures that the recommendation system always has fresh data about current consumer interests, replacing the periodic or batch processing of stale historical data

Inventive Principle:

Principle #20Continuity of useful action

Solution Approach 2:

The patent establishes a feedback loop where consumer interactions with tagged regions are immediately captured, analyzed, and used to update recommendations. This feedback mechanism ensures that the system continuously learns from current consumer behavior rather than relying on outdated historical patterns, maintaining data freshness and relevance

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS10366440B2Monitoring consumer-product view interaction to improve upsell recommendations

Publication Date: 2019.07.30 ADOBE INC

AI summary

In various implementations, a visual representation of a product is presented to an online consumer. The visual representation can have one or more regions that are each associated with one or more feature tags associated with a particular category of the product. While the consumer interacts with the visual representation, interaction times between the consumer and the product's visual representation are monitored and measured to determine which features of the product appear to be of particular interest to the consumer. Based on the monitored interaction times, product upsell recommendations associated with the same category of the product are generated and presented to the user.