See Your Data's Family Tree No Guessing k

The method, in full

The practical guide behind this tool: when it applies, how to read the output, and the traps that make it say the wrong thing.

Read the guide →

How it works

The analysis standardizes your selected columns, computes euclidean distances, and builds the full agglomerative cluster tree under Ward.D2 linkage, merging the two closest groups over and over until one remains. Complete and average linkage trees are also built and all three are scored by cophenetic correlation (how faithfully each tree preserves the original distances); the best is reported, while the reported clusters are cut from the Ward.D2 tree for its compact, interpretable groups. The cut point is chosen by average silhouette width over every candidate k, cross-checked against the tree's own merge-height gaps, and each segment is profiled and illustrated with named example rows.

Use it when you suspect natural groups but don't want to pre-pick how many (the tree shows structure at every scale, including sub-segments nested inside segments) or when you want to see exactly how close any two groups are to each other.

Not for categorical-only data or very large tables (the distance matrix caps the analysis at a 1,500-row sample: use the k-means tool for bulk segmentation), and not when clusters are elongated or density-defined shapes.

What data do you need?

Any table where each row is a named entity described by numeric columns. For example, customer behavior metrics:

Minimum 30 rows · Best with 100-1,500 rows and 2-12 numeric columns (larger tables run on a reproducible 1,500-row sample)

What's in the report?

Standard-library analysis: agglomerative hierarchical clustering on standardized features. Unlike k-means, the full cluster tree is built first, structure at every scale, and only then cut, with the number of segments chosen by average silhouette width and cross-checked against the tree's own merge-height gaps. Ward.D2 linkage for compact interpretable groups, three linkages compared by cophenetic correlation, profiles as a z-score heatmap, a 2-D map, and every segment illustrated with named example rows. Works on any dataset: map 2 or more numeric columns plus a label column.