Cluster analysis is a family of statistical methods that sort observations into groups so that members of the same group are more alike than members of different groups. In market research it is the engine behind segmentation: instead of treating a market as one undifferentiated crowd, you let the data reveal the natural groupings inside it, then build strategy around them.
What it is good for
Segmentation answers a practical question: who are our customers, really, and how do they differ? Done well, it produces segments that are distinct, sizeable enough to matter, and reachable through real channels. That lets a team tailor products, messaging, and pricing to groups that actually behave differently, rather than to an average customer who does not exist.
How the common methods differ
- k-means: fast and widely used; you choose the number of segments and the method finds compact groups. Best for large, numeric datasets.
- Hierarchical clustering: builds a tree of nested groups, so you can see structure at several levels before committing to a number.
- Latent class / model-based clustering: treats segment membership as a probability, which handles categorical survey data and uncertainty more honestly.
Cluster analysis vs factor analysis
They are often confused. Factor analysis groups variables that move together, reducing many survey questions to a few underlying themes. Cluster analysis groups respondents. In practice the two work together: factor analysis first, to distil the questions into meaningful dimensions, then cluster analysis on those dimensions to form the segments.
Where it goes wrong
The number of segments is a judgement, not a fact the software hands you; picking it well takes both statistics and business sense. Variables measured on different scales need to be standardised first, or one will dominate. And a segmentation is only useful if it is stable and interpretable, a solution that looks tidy but cannot be described to a marketing team has failed, however good the fit statistics look.
If you have survey or behavioural data and need segments your stakeholders can act on, that is our work. See our market-research analysis or tell us about your data.