Advanced Methodologies

Hierarchical clustering

Build a nested clustering tree to explore how cases group together before selecting a solution.

  • Segmentation
  • Clustering

What it is

Hierarchical clustering progressively merges or splits observations to reveal a nested grouping structure often visualized with a dendrogram.

Overview notes

Useful role

This method is often best used to understand the structure of the problem before a final operational clustering step.

Decision guide

When to use it

  • When you want to explore clustering structure before fixing a final solution
  • When a dendrogram helps explain grouping logic
  • When cluster count is still uncertain

When not to use it

  • When the sample is very large and speed matters
  • When you need a quick production-ready assignment method

Inputs required

  • Standardized variables or a valid distance matrix
  • A linkage choice

Typical outputs

  • Dendrogram
  • Candidate cluster solutions
Simple example

Examine how respondents cluster on attitude statements to decide whether a three four or five cluster solution is more defensible.

Strengths
  • Exploratory and visually intuitive
  • Helpful for judging cluster count
Limitations
  • Can be computationally heavier
  • Sensitive to distance and linkage choices
Common mistakes
  • Treating one dendrogram cut as obviously correct
  • Ignoring the effect of preprocessing and linkage selection
How I use it in practice

I use hierarchical clustering when the team still needs to understand the structure of the data before locking a final segmentation solution.

What is outputted
  • Dendrogram and candidate cluster cuts
How to interpret the output
  • Look for stable interpretable branches rather than forced symmetry
How to communicate to clients
  • Explain what the dendrogram shows and what it does not prove
Displayr / Q implementation notes
  • Preserve the distance and linkage settings in the project notes

Visual placeholder

Dendrogram placeholder

Add a screenshot of a dendrogram and annotate where a plausible cut might happen.

Recommended placeholder: chart screenshot, process diagram, output interpretation notes, and one short caption on what to inspect first.

Related topics

Jump to connected concepts, techniques, or implementation notes.