COMPEL Glossary / clustering
Clustering
Clustering is an unsupervised learning technique that groups similar data points together based on shared characteristics, without requiring pre-labeled categories.
What this means in practice
Common applications include customer segmentation (discovering distinct buying patterns), document categorization, fraud ring detection, and market analysis. A clustering model might analyze purchasing behavior across millions of customers and identify five distinct segments that no human analyst had previously recognized. For organizations, clustering is valuable for exploration and discovery -- it reveals structure in data that can inform business strategy, marketing campaigns, and targeted interventions. Clustering results often become the foundation for subsequent supervised learning projects that operationalize the discovered patterns.
Why it matters
Clustering reveals structure in data that human analysts may never discover, identifying natural groupings that can inform business strategy, marketing campaigns, and targeted interventions. Its value lies in exploration and discovery — finding patterns that no one was looking for. Organizations that use clustering effectively gain customer insights, detect fraud rings, and discover market segments that provide competitive advantage through deeper understanding of their data.
How COMPEL uses it
During the Calibrate stage, clustering may be used as an analytical technique to discover patterns in organizational data as part of the data readiness assessment. The Model stage evaluates clustering use cases for portfolio inclusion. The Produce stage develops and deploys clustering models, with results often becoming the foundation for subsequent supervised learning projects. The Evaluate stage measures whether discovered clusters translate into actionable business insights.
Related Terms
Other glossary terms mentioned in this entry's definition and context.