KPI-Driven Customer Segmentation Using Predictive Topic Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing analytics systems fail to generate actionable customer segments tied to specific business outcomes, lacking correlation with desired results and providing insufficient insights for decision-making.
Innovation Solution
A KPI analytics system that combines predictive modeling with unsupervised learning to identify variables affecting outcomes, using propensity models and topic modeling to segment customers based on predicted outcomes, providing actionable insights for targeted campaigns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If conventional customer segmentation methods are used, then customer groups can be identified, but the segments lack actionable variables that explain why outcomes occur
Solution Approach 1:
The patent segments customers into distinct groups based on predicted outcomes and the variables that drive those outcomes. Each segment is characterized by specific actionable variables that explain why certain outcomes occur, allowing businesses to understand not just who belongs to each segment but why they exhibit certain behaviors or outcomes.
Solution Approach 2:
The patent introduces an intermediary analytical layer that connects customer data to outcomes through identified variables. This intermediary process analyzes the relationship between customer characteristics and outcomes, extracting actionable variables that serve as mediators explaining the causal relationships, thus preventing loss of explanatory information.
2Measurement precision
If predictive modeling is used to identify outcome correlations, then outcome prediction accuracy improves, but the system fails to provide actionable variables for decision-making
Solution Approach 1:
The patent segments customers not just by outcome probability but by the specific variables that drive those outcomes. This segmentation approach maintains prediction accuracy while simultaneously providing actionable insights by grouping customers based on the actionable variables that explain their predicted outcomes.
Solution Approach 2:
The patent applies local quality by providing different actionable variable information for different customer segments. Each segment is characterized by specific variables that are locally relevant to that group's predicted outcomes, allowing decision-makers to understand what actions to take for each specific segment rather than losing actionable information.
3Adaptability or versatility
If traditional analytics systems provide customer segments, then customer grouping is achieved, but the segments are not tied to specific business outcomes or KPIs
Solution Approach 1:
The patent creates a universal segmentation framework that can be applied to any business outcome or KPI. The system is designed to work with different types of outcomes (churn, conversion, revenue, etc.) and consistently provides actionable variables tied to those specific outcomes, making the segmentation versatile while maintaining strong outcome correlation.
Solution Approach 2:
The patent incorporates feedback by continuously analyzing the relationship between customer variables and outcomes. The system uses outcome data to refine and validate the actionable variables identified for each segment, creating a feedback loop that ensures segments remain tightly correlated with specific business outcomes while maintaining versatility across different KPIs.
Data Source
AI summary
An improved analytics system generates actionable KPI-based customer segments. The analytics system determines predicted outcomes for a key performance indicator (KPI) of interest and a contribution value for each variable indicating an extent to which each variable contributes to predicted outcomes. Topics are generated by applying a topic model to the contribution values for the variables. Each topic comprises a group of variables with a contribution level for each variable that indicates the importance of each variable to the topic. User segments are generated by assigning each user to a topic based on attribution levels output by the topic model.