Tick density should match the precision readers need

What it is

Ticks are the only official channel for converting mark position back into numbers, and their density sets the readout precision this chart supports: on an axis with 20-unit intervals, the best a reader gets is "a little past 40"; if the question is "37 or 38?", the axis cannot answer. Too sparse, and reading becomes guessing; too dense, and tick labels crowd each other while the search for the nearest tick slows down — the axis turns into a noise band. Density has exactly one determinant — what precision of question readers will ask of this chart — not "whatever the library defaults to."

Why it happens

The readout action is "target position → nearest tick → tick value ± half-interval interpolation": the precision ceiling = half the tick interval (half-interval is the best possible estimate). The interval must sit at least one notch below the question's required precision for stable reading (reading at granularity 5 needs intervals ≤ 2); coarser than that, readers interpolate on their own between ticks with error multiplied. The cost of excess density is equally real: each added tier raises the axis's visual weight, and finding the nearest tick goes from "at a glance" to "one by one," dragging glance-style reading (trend comparison) — which is why the major-tick/minor-tick hierarchy is the compromise between density and order.

Where it stops holding

Precision needs are set jointly by question and audience: operational readouts (numbers to be transcribed into reports) need fine ticks or direct labels; executive scanning ("up or down?") is served by two or three tiers. Interactive hover readouts offload "exact values" from the axis entirely (hover delivers full precision), letting axis ticks retreat to a coarse orienting function — a key fork between static and interactive charts. Log axes place density by decade (one major tick per order of magnitude) with minor ticks at 2× and 5× — the same logic, though intervals are unequal in raw-value space.

Applying it

• Write the question's precision requirement ("read to hundreds") before fixing ticks; convert to an interval at most half that; beyond five major tiers, switch on the major/minor hierarchy.
• Interactive charts delegate precise readout to hover, keeping three to five "orientation ticks"; static report charts go the other way — denser ticks or direct labelling.
• Verification: have readers read designated points and compare with truth; the error distribution's width is the actual precision — compare with the question's need, thinning or densifying accordingly.

Related

• Same group: U3.05.2 Gridlines must stay weaker than data elements · U3.05.3 Direct labelling can replace gridlines and legends
• Nearby: U2.15.5 Small multiples compare shapes, not precise values · U2.16.1 Tables beat charts when exact values are needed
• Search terms: tick density · readout precision · axis ticks · major minor ticks