Treemaps suit magnitude composition in hierarchies

What it is

When relations form a strict parent-child hierarchy and the main question is how much each branch contributes to a whole, a treemap is more direct than a general node-link diagram: nested rectangles encode containment and area encodes nonnegative magnitude. The data need a documented, single-parent rooted hierarchy or a reduction to one without material loss, and values must form an interpretable composition under each parent. If readers must trace paths, inspect cross-branch edges, or understand nonhierarchical relations, area composition cannot replace a network view.

Why it happens

Containment maps parent-child relations to spatial nesting, while area allocates each parent's fixed region among its children. Readers can locate large branches within a local whole and drill down without a field of edges. Compared with a node-link tree, the treemap uses planar space efficiently but removes explicit paths and adjacency. When leaves use additive values in one unit and a global area scale, cross-branch leaf areas remain geometrically proportional to raw values, although nesting hinders direct judgment. Local shares under different parents have different denominators, and parent-child or cross-level nodes are not independent comparison items.

Studying it

Compare a treemap with a node-link tree, indented table, and hierarchical bar view using the same hierarchy and values. Separate largest-branch, share-estimation, ancestor-lookup, path-tracing, cross-branch comparison, and cross-link discovery tasks; record error, time, and navigation loss. Manipulate depth, branching, skew, label length, and screen size, and test whether readers identify the denominator. In realistic work, check whether aggregation, duplicate counting, or unknown parents changes the composition claim.

Where it stops holding

Multiple parents, cycles, and important cross-links cannot be reduced to a tree without loss; choosing one parent hides relations. Negative values, overlapping categories, nonadditive measures, and mixed units do not form nested area shares. If the root is not a complete and stable whole, a local percentage cannot be generalized globally. “Suits” does not imply precise comparison: sorting, tables, or bars may better serve ranks and read-off. The term treemap must also remain distinct from a node-link tree diagram.

Applying it

• Validate each node's parent, aggregation, unit, and denominator; expose unassigned values, duplicates, and residuals.
• Use breadcrumbs, stable labels, and reversible drill-down to communicate the current root and hierarchy, preventing local shares from being read globally.
• Add a table or common-scale view for exact values, ranks, and cross-branch comparisons; use node-link or matrix views for cross-links.
• Test conservation, ancestor lookup, accessible labels, and keyboard or screen-reader navigation with representative data instead of accepting a tidy layout as success.

Related

• Same group: U2.13.2 Layout algorithms shape node-link readings · U2.13.3 Dense networks collapse into hairballs
• Adjacent: U2.11.1 Treemaps encode shares as areas of nested rectangles · U2.11.5 Treemaps cannot show negatives or support cross-branch comparison
• Search terms: treemap · hierarchical composition · part-to-whole · tree versus network