Description
Groups users or observations into homogeneous segments based on similarity across variables, used for behavioural segmentation and data-driven personas.
Cluster analysis is a family of unsupervised techniques that partition observations into groups whose members are more similar to each other than to members of other groups. Common methods include k-means, hierarchical clustering, Gaussian mixture models and density-based approaches. In UX it supports segmentation from behavioural logs or survey responses and provides quantitative grounding for personas. Results depend heavily on choices: which variables, how they are scaled, the distance measure, the algorithm and the number of clusters. Rigour comes from theory-driven variable selection, standardisation, comparing solutions with internal validity indices such as silhouette scores, testing stability with resampling or split halves, and checking that clusters differ meaningfully on external variables not used in clustering.
Attributes
- Discipline
- Statistics & Modelling
- Loss type
- Active
- Goal
- Segment users
- Stage
- Definition, Discovery
- Data
- Quantitative
- Effort
- Weeks
Steps
- Define the purpose of segmentation and select variables that relate to it.
- Clean data, handle missing values and standardise variables.
- Reduce dimensionality or remove highly correlated variables if needed.
- Run one or more algorithms and compare solutions with different numbers of clusters using silhouette, elbow or information criteria.
- Test stability by re-running on random subsamples or split halves.
- Profile clusters on clustering and external variables and name them.
- Validate usefulness with stakeholders and qualitative research before building personas or targeting.
Sources
- Cluster Analysis for Researchers
- scikit-learn
- Cluster Analysis (Everitt et al.)
- Factor and Discriminant Analysis (Sauro & Lewis, MeasuringU training)
- Practical Statistics for Data Scientists (Bruce & Bruce)
- Myatt, G., Making Sense of Data: A Practical Guide to Exploratory Data Analysis and Data Mining (2007)
- Drachen, A., Mirza-Babaei, P. & Nacke, L. (eds), Games User Research (2018)
- Veal, A. J. & Darcy, S., Research Methods in Sport Studies and Sport Management (2014)
Also known as
- Clustering
- Segmentation analysis
- k-means
- Hierarchical clustering
Categories
- Analytics & Behaviour
- Statistics & Modelling