A cluster marker drawn the same size regardless of count teaches readers a false sense of uniformity

What it is

Once points are folded into cluster markers, the only channels left for "how many points are in here" are the marker's size and its numeric label. If every cluster is drawn the same size, readers learn that clusters exist but cannot judge their relative weight—and the aggregation has thrown away the one piece of information it deliberately kept. Cluster marker size must map the contained count by area, with numeric labels carrying the precise reading.

Why it happens

Folding discards geometry but intentionally preserves the total, and that total needs a visual outlet. Size is the natural quantity channel on a cluster marker, but the mapping decides whether the reading is correct. Perceived symbol magnitude follows area, so mapping the count linearly to radius or diameter compresses differences: a 10x difference in count appears as only about a 3.2x difference in diameter, and readers underestimate how much a large cluster outweighs a small one. Mapping by area (area proportional to count) aligns perceptual weight with numeric weight. The channel's capacity is finite: only a limited number of area levels can be told apart reliably, and with extreme ranges the largest cluster flattens the visual differences among all the others—at that point the numeric label becomes the primary reading channel and size degrades to a coarse cue.

Where it stops holding

Size encoding works best when there is a moderate number of clusters and the magnitude range stays within about two orders. Heavy-tailed data—one giant cluster plus many small ones—defeats even area mapping, and labels carry the reading. When many clusters sit close together, size differences drown in the crowding. And when size is combined with other channels such as color intensity or opacity, readers do not know which channel to read and tend to attend to only one.

Applying it

• Scale cluster marker area proportionally to the contained count; do not map counts linearly to radius or diameter.
• Label the largest, smallest, and mid-range clusters with their counts, covering the intervals where size alone reads poorly.
• Verify by asking readers to rank several clusters by weight from the map alone, then compare against true counts; misranked pairs expose where the size mapping distorts.

Related

• Same group: U12.03.1 Point maps systematically undercount in high-density regions because of overplotting · U12.03.2 Aggregation into clusters hides within-cluster distribution differences · U12.03.4 Hexagonal binning or density surfaces can replace direct point overlay · U12.03.5 Zoom-dependent aggregation radii make the same data support different conclusions
• Nearby: U12.04.2 Symbols must map value to area and carry a size legend · U1.09.1 Each visual channel supports only a limited number of reliably distinguishable levels
• Search terms: cluster size · proportional symbol · area encoding