Session-Specific Recommendation Personalization via Real-Time Signals

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Ecommerce platforms struggle to provide personalized recommendations to new customers as existing systems rely on historical data, which is not available for new users, leading to generic recommendations that do not adapt quickly to individual user preferences.

Innovation Solution

A session-specific conversion determination system that uses real-time signals from user interactions to generate probability affinities for item facets, employing a machine learning model with interaction parameters and historical data to personalize recommendations in real-time.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If historical data is used for recommendations, then personalization accuracy is improved, but new customers cannot receive personalized recommendations because they lack historical data

Engineering Contradiction:

Improvepersonalization accuracyVSAvoidapplicability to new customers

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system performs preliminary actions by collecting and analyzing real-time interaction signals during the user session to build a personalized profile before the user completes their shopping journey. This allows the system to proactively personalize recommendations for new customers based on their current session behavior rather than waiting for historical data to accumulate.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces real-time interaction signals as an intermediary element that bridges the gap between new customers without historical data and the personalization engine. These signals serve as a temporary substitute for traditional historical data, enabling the machine learning model to generate personalized recommendations for users who would otherwise be ineligible.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Device complexity

If generic recommendations are provided to new customers, then system simplicity is maintained, but user experience deteriorates due to lack of personalization

Engineering Contradiction:

Improvesystem complexityVSAvoiduser experience

Core Design Contradiction:

Device complexityVSEase of operation

Solution Approach 1:

The recommendation system transitions from a static generic approach to a dynamic personalized approach by continuously processing real-time interaction signals during the user session. The system adapts its recommendations based on evolving user behavior within the session, making the user experience more responsive and personalized without requiring complex pre-processing of historical data.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent changes the input parameters from traditional historical user profiles to real-time interaction signals. This parameter transformation enables personalization for new customers by using current session behavior (clicks, views, time spent) as the basis for recommendations, thereby improving user experience while maintaining system feasibility.

Inventive Principle:

Principle #35Parameter changes

3Speed

If real-time signals are processed to personalize recommendations, then personalization speed is improved, but computational requirements increase

Engineering Contradiction:

Improvepersonalization speedVSAvoidcomputational resources

Core Design Contradiction:

SpeedVSUse of energy by moving object

Solution Approach 1:

The system extracts only the most relevant features from real-time interaction signals rather than processing complete user profiles. By focusing on key interaction parameters (item views, clicks, time spent on pages) during the current session, the system reduces computational overhead while maintaining personalization effectiveness and speed.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies partial action by using a subset of available interaction data sufficient for effective personalization rather than analyzing all possible user behaviors. This selective approach enables fast personalization for new customers by processing only the critical real-time signals needed to generate accurate recommendations without exhaustive computation.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS12211082B2Systems and methods for personalizing recommendations using real-time signals of a user session

Publication Date: 2025.01.28 WALMART APOLLO LLC

AI summary

A session-specific conversion determination system can include a computing device configured to receive real-time signals of an event occurring from a user device. The real-time signals include interaction parameters. The computing device is also configured to obtain a set of historical data based on the interaction parameters and a set of facets and generate a probability affinity for each facet of the set of facets by implementing a machine learning model using the interaction parameters and the set of historical data as features. The computing device is also configured to adjust a display of a set of recommended items based on the probability affinity for each facet of the set of facets and transmit the display of the set of recommended items to a user interface of the user device.