Direct labelling can replace gridlines and legends

What it is

The conventional reading path puts information far away: values on the axis, names in the legend, and the reader's eyes shuttle among data, axis, and legend. Direct labelling moves both to where the eyes already are — series names at the line ends (replacing the legend), values beside key points (replacing gridlines and ticks). The shuttling stops, working memory no longer holds "which one is blue," and the reading chain shortens from "trace line - consult legend - back to line - align with grid - estimate" to "what you see is what it is." The chart can then go cleaner: gridlines withdrawn, legend deleted, and every remaining element working.

Why it happens

The economy comes from fixation distance: legend reading costs two saccades plus one colour-name binding (an error-prone operation that collapses past four colours); a line-end label puts the name beside the line's terminus, binding completed in one fixation and immune to colour confusion (position binding is sturdier than colour binding). Value labels work the same way: grid reading is position-to-tick interpolation (with error), while a digit beside the point is zero-error readout. Direct labelling also buys selectivity — only key points (peaks, valleys, latest, anomalies) are labelled, concentrating the chart's precision resources where readers actually pause, instead of spreading them evenly across a grid.

Where it stops holding

Labels are a scarce resource and collide when abundant: beyond three or four fully labelled series, line ends crowd, and the design retreats to "legend plus key-point labels"; with dense points, label only anomalies and extremes. Labels also occlude when competing for space (end labels covering the final segment's movement), calling for avoidance rules (offset, leader lines, vertical staggering). Interactive hover readout is the dynamic form of direct labelling; static media must pre-decide "where to label" — the two forms coexist, reports often pairing "static key labels with interactive hover for everything." What direct labelling replaces is the grid-and-legend's readout and identification duties; the axis itself stays (scale declaration, units, sense of magnitude).

Applying it

• Default line charts to end-of-line series labels (text in the line's colour), invoking the legend only for many-series cases; label peaks, valleys, and latest values directly, thinning or removing the grid accordingly.
• Avoidance spec: offset labels one or two character-widths from the line end, stagger vertically on overlap, short leader lines where needed; label type never smaller than tick type.
• Verification: compare task times (name the series, read a key value) between "legend + grid" and "direct label" versions; the labelled version should be markedly faster with series-name confusion errors gone.

Related

• Same group: U3.05.1 Tick density should match the precision readers need · U3.05.2 Gridlines must stay weaker than data elements
• Nearby: U1.13.3 Common region can replace the legend for belonging · U1.11.2 Encoding that jumps between charts forces readers to relearn the legend
• Search terms: direct labelling · line-end labels · annotation · legend replacement