Scaling Consumer Intelligence with Google Cloud

Situation

A leading home improvement retailer wanted to improve how it understood and engaged its large Consumer customer base. While an effective segmentation framework already existed for Pro customers, Consumer customers were largely treated as a single audience, limiting personalization opportunities across marketing, merchandising and digital channels.

The Customer Insights team required a scalable cloud-native solution capable of analyzing millions of customers, identifying meaningful behavioral segments and operationalizing those insights across enterprise applications through automated refreshes.

Problem

• Consumer customers were addressed as a single segment despite highly diverse shopping behaviors
• Existing segmentation relied primarily on basic demographic information with limited behavioral intelligence
• Manual segmentation approaches could not scale across more than 56 million consumers
• Business teams lacked a common segmentation framework and consistent customer language
• Regular segment refreshes required significant manual effort and long processing times
• Marketing and merchandising teams needed actionable customer insights to improve personalization and campaign effectiveness

Solution

• Built an end-to-end consumer segmentation platform on Google Cloud using behavioral, transactional and channel data
• Engineered approximately 50 customer features from an Analytical Data Set (ADS) containing over 56 million known consumers
• Applied advanced preprocessing techniques including Signed Log Transformation, Standard Scaling and One-Hot Encoding to prepare data for machine learning
• Evaluated multiple clustering algorithms including K-Means, Gaussian Mixture Models (GMM), Hierarchical Clustering and DBSCAN, selecting K-Means based on business interpretability and operational efficiency
• Developed five actionable consumer segments representing distinct shopping behaviors and channel preferences
• Trained an XGBoost classifier using K-Means cluster assignments, enabling fast and scalable weekly prediction of consumer segments without retraining the clustering model
• Automated weekly model execution, data validation and dashboard refreshes through cloud-native orchestration

Key Components:

Scalable Customer Data Platform: BigQuery served as the centralized analytical data warehouse, processing transactional, customer, channel and behavioral data for over 56 million consumers while supporting feature engineering and model execution.

Advanced Customer Segmentation: Vertex AI and BigQuery ML enabled experimentation with multiple clustering algorithms and deployment of the selected K-Means model to identify behaviorally distinct customer segments.

Efficient Weekly Scoring: An XGBoost classification model was trained on K-Means outputs, allowing rapid weekly prediction of consumer segments using the latest customer data while maintaining segmentation consistency.

Automated Orchestration: Cloud Composer automated feature engineering, model execution, data quality validation and scheduled weekly refreshes, ensuring reliable production operations.

Centralized Data Management: Cloud Storage maintained model artifacts, intermediate datasets and historical outputs while IAM governed secure enterprise access.

Business Intelligence & Activation: Consumer segments were published through the Know Your Consumer (KYC) portal using Looker and integrated with downstream marketing and CRM systems through Cloud Run APIs, enabling enterprise-wide adoption.

Decision Science Framework: Mu Sigma’s structured problem-solving approach – spanning problem definition, hypothesis framing and decision architecture – ensured segmentation outputs were designed for enterprise action, not just analytical reporting. This connected consumer segments directly to measurable business decisions across teams, going beyond conventional modeling engagements.

Together, Google Cloud services transformed customer segmentation from a manual analytical exercise into an automated AI-powered intelligence platform that continuously delivers actionable customer insights across the organization.

Impact

• Segmented more than 56 million known consumers into five actionable behavioral segments
• Top-performing consumer segment contributed approximately 27% of R12 sales
• Automated weekly segmentation refreshes, eliminating manual operational effort
• Enabled enterprise-wide access to standardized consumer segments through the KYC portal and downstream marketing systems