Data-driven customer segmentation using vector clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current customer segmentation methods in retail are largely manual, rely on predefined rules, and produce sub-optimal results due to high variance within segments, leading to lost sales and inefficiencies, as they are not dynamic and require significant analytics and time for manual analysis.
Innovation Solution
A data-driven approach that transforms item codes into multidimensional vectors using Word2Vec algorithms, aggregates consumer-item vectors from transaction histories, and applies clustering algorithms to dynamically create customer segments, eliminating the need for predefined segments and manual rules, allowing for real-time adjustments based on changing purchase behaviors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If manual rules-based segmentation is used, then predefined segments can be created, but the variance between segment members is high and results are sub-optimal
Solution Approach 1:
The patent replaces manual mechanical segmentation processes with automated machine learning algorithms. Specifically, it uses clustering algorithms (such as K-means, hierarchical clustering, or DBSCAN) that automatically analyze transaction data and identify customer segments based on purchasing patterns, replacing the need for manual rule creation and application.
Solution Approach 2:
The patent transforms the segmentation approach by changing from fixed predefined parameters to dynamic data-driven parameters. It extracts features from transaction histories (such as purchase frequency, basket composition, spending patterns) and uses these as input parameters for clustering algorithms, allowing segments to be defined by actual behavioral patterns rather than static rules.
2Measurement precision
If manual analysis is used to define segments, then segments can be created, but substantial analytics time must be invested by analysts
Solution Approach 1:
The patent implements self-service segmentation where the system automatically analyzes transaction data and creates segments without requiring manual intervention from analysts. The machine learning algorithms independently process customer transaction histories, identify patterns, and generate segments autonomously, eliminating the need for substantial manual analytical time.
Solution Approach 2:
The patent performs preliminary data processing and feature extraction automatically before segmentation. It pre-processes transaction histories to extract meaningful features (such as normalized purchase patterns, basket composition ratios, temporal patterns) that serve as input for clustering algorithms, eliminating the need for manual data preparation and analysis.
3Ease of manufacture
If predefined segments are used, then segmentation can be performed, but the segments are static and not dynamic
Solution Approach 1:
The patent transforms static predefined segments into dynamic data-driven segments that automatically adapt to changing customer behaviors. The clustering algorithms continuously process new transaction data, allowing segments to evolve and reconfigure themselves based on current purchasing patterns, ensuring they remain relevant and accurate over time.
Solution Approach 2:
The patent implements feedback mechanisms where segmentation results are continuously validated and refined through ongoing analysis of transaction data. The system monitors segment performance and automatically adjusts segment definitions and assignments based on new data, creating a closed-loop system that adapts to changing customer behaviors and market conditions.
4Ease of operation
If rules-based approach is used, then customer assignment to segments can be done, but specific rules must be defined and implemented requiring programmer assistance
Solution Approach 1:
The patent replaces manual rule definition and implementation with automated machine learning algorithms. Instead of requiring programmers to encode complex segmentation rules, the system uses clustering algorithms that automatically learn optimal segmentation criteria from transaction data, eliminating the need for manual rule programming while simplifying the overall system complexity.
Data Source
AI summary
Item codes are mapped to multidimensional space as item vectors based on each item codes context relevant to other item codes in a product catalogue. A transaction history for a given customer is obtained and each item vector associated with a corresponding item purchase made by that customer is obtained. All item vectors per customer are summed to create an aggregated and single vector representing the purchase history of each customer. The aggregated customer-item vectors for the customers are plotted in the multidimensional space. The plotted customer-item vectors are then clustered into groupings based on their distances from one another in the multidimensional space; the groupings representing data-driven customer segments. The data-driven customer segments along with customer identifiers for the customers comprising each segment are provided as input to promotional engines and/or loyalty systems.