Shrinking the marks to relieve overcrowding just hides the problem instead of actually solving it

What it is

Faced with a density overflow, the wrong reflex is "make the points smaller, dial down the opacity" — the canvas looks relieved while the problem stands untouched: the data volume is unchanged, the occlusion relations are unchanged, only the visibility of the occlusion has dropped. The correct place to decide is the data-processing stage: aggregate, sample, or facet — all data transformations, settled before drawing. Mark size serves readability and nothing else; it must never be conscripted into density governance.

Why it happens

Shrinking marks is a disguised fix because it acts on the perception layer while the problem lives in the information layer: smaller marks leave every overlap intact but spend fewer pixels per overlap, making it harder for readers to discover that "multiple points live here" at all — occlusion converts from a visible problem into a hidden one, and misreading deepens. Transparency stacking is the same story: dark regions do hint at density, but without a scale there is no quantity, so readers cannot tell how many points stacked. Data-processing remedies change the information structure itself — reducing the information that must be shown at once, or turning density into an explicit encoding. That is the cure.

Where it stops holding

Mark size and opacity are not banned — they are legitimate auxiliaries: on an already-aggregated or sampled chart, smaller marks and alpha blending fine-tune readability, provided the information-structure problem is solved first; what is banned is using them as a substitute for processing. A second boundary is rendering performance: with truly massive data, drawing itself fails — cutting the data volume first serves both readability and engineering, and the two reasons point at the same action.

Applying it

• Set the process rule: the density review happens in data processing; decide aggregate/sample/facet before drawing, and fix mark styling last.
• In review, when a chart shows "points too small to see," ask about data volume and processing before discussing style.
• Verification: inspect the implementation of high-density charts — aggregation or sampling logic present with normal-sized marks means compliant; brute-forcing with miniature marks goes back for rework.

Related

• Same group: U5.06.1 One chart's mark count is limited by overlap and occlusion · U5.06.2 Past the density ceiling, aggregate, sample, or facet · U5.06.3 The density ceiling shifts with final size and resolution · U5.06.4 High-density charts keep overall shape and lose individual readability
• Nearby: U5.06.2 Past the density ceiling, aggregate, sample, or facet · U5.05.3 De-emphasised elements keep a readable minimum contrast
• Search terms: overplotting remedies · data reduction · alpha blending limits