E-Commerce Return Propensity Modeling With Contextual ML Signals

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional methods lack systematic analysis of key driving forces behind consumer product returns in e-commerce, leading to inefficiencies and substantial losses for retailers and manufacturers, with existing approaches focusing on predicting return policies based on customer behaviors or product categories rather than individual consumption patterns.

Innovation Solution

A method and system utilizing multivariate machine learning models to analyze historical transaction data, including return spreads, similarity spreads, and linkage spreads, to predict potential returns by mapping return patterns with customer profiles and contextual information, adjusting cut-off values for enhanced accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional return prediction methods based on customer behaviors or product categories are used, then implementation simplicity is maintained, but prediction accuracy of potential returns deteriorates

Engineering Contradiction:
Improveprediction accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments return prediction into multiple dimensions: customer behavior patterns, product category characteristics, and contextual factors (local events, seasonal variations). Each dimension is analyzed separately through multivariate machine learning models, allowing comprehensive prediction while maintaining manageable complexity through modular analysis of distinct data segments

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transitions from traditional one-dimensional return prediction (based solely on customer behavior or product category) to multi-dimensional analysis by incorporating contextual information such as local events, seasonal patterns, and environmental factors. This dimensional expansion enables more accurate prediction by capturing the complex interplay of multiple influencing factors

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Measurement precision

If multivariate machine learning models with multiple data dimensions are implemented, then prediction accuracy of potential returns is improved, but computational complexity and data processing requirements worsen

Engineering Contradiction:
Improvereturn propensity estimation accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent performs preliminary data processing and feature extraction before main prediction analysis. Historical transaction data is preprocessed to identify relevant patterns, and contextual information is pre-filtered to extract only significant factors. This preliminary action reduces the dimensionality of input data, lowering computational complexity while preserving prediction accuracy

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent dynamically adjusts model parameters and cut-off values based on the specific characteristics of the data being analyzed. By changing parameters adaptively rather than using fixed thresholds, the system optimizes the balance between model complexity and prediction accuracy for different product categories and customer segments

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If dynamic adjustment of cut-off values is implemented, then return propensity estimation accuracy is maximized, but processing time and computational resources worsen

Engineering Contradiction:
Improvereturn propensity estimation accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent implements periodic adjustment of cut-off values rather than continuous dynamic adjustment. Cut-off values are updated at predetermined intervals or when significant changes in data patterns are detected, rather than being recalculated for every prediction. This periodic approach maintains high accuracy while significantly reducing processing time and computational resource requirements

Inventive Principle:
Principle #19Periodic action

Data Source

PatentEP4617971B1Method and system for managing item returns
Publication Date: 2025.12.31 TATA CONSULTANCY SERVICES LTD
  • EP4617971B1 patent drawingFigure 1A
  • EP4617971B1 patent drawingFigure 1B
  • EP4617971B1 patent drawingFigure 2A

AI summary

The present disclosure estimates items to be returned based on nature of items picked online in association with the context. The context includes time, shopper details, local events and the like. It is addressed by mapping return spread and similarity spread or linkage spread in unique format. Further, the intrinsic mechanism that result returns are captured by training a multivariate Machine Learning (ML) model using the actual return spread, the similarity spread, or the linkage spread and the customer profile data. The captured return mechanism is leveraged to pre-empt the returns online in the form of return spread at the time of ordering in real time.