Principles
This reference defines the judgment standard for quantitative displays. Use display-selection.md for chart-type choices and integrity-audit.md for distortion checks.
Graphical excellence
A strong quantitative display communicates complex evidence with clarity, precision, and efficiency. It helps viewers compare, inspect, and reason more effectively than the raw table alone, while preserving enough context to judge the evidence honestly.
Good displays often do several things at once:
- reveal comparisons, variation, outliers, clusters, trends, uncertainty, or spatial structure;
- align visual magnitude with numerical magnitude;
- integrate words, numbers, and pictures close to the evidence;
- support both overview and detail when the medium allows;
- organize complexity rather than deleting useful complexity.
Evidence before decoration
Ask what each mark does. A mark earns its place when it encodes data, labels data, supports comparison, shows context, explains uncertainty, improves accessibility, or guides the viewer through a difficult comparison.
A mark is suspect when it mainly decorates, dramatizes, repeats information, or pulls attention away from the evidence. This is not a ban on all non-data ink: light gridlines, axes, source notes, captions, thresholds, uncertainty legends, panel labels, and annotations are often essential.
Integrity before elegance
A calm-looking chart can still be dishonest. Check whether scales, baselines, transformations, encodings, denominators, time windows, filters, and uncertainty support the claim. Do not rely on printed numbers to rescue a misleading visual encoding.
Comparison is the central act
Most visualization problems are comparison problems. Improve a display by making the intended comparison easier: align common baselines, sort deliberately, use common scales, facet repeated structures, label directly, add meaningful reference lines, or show paired differences.
Density is not clutter
Low-density charts can waste attention and hide context. High-density displays can be clear when they are ordered, consistently scaled, annotated, and visually quiet. Remove useless marks before removing useful data.
Useful density can come from raw observations, small multiples, compact points, context bands, reference series, marginal distributions, or carefully integrated labels. Do not inflate a few numbers into a grand graphic when a sentence or small table would be clearer.
Words, numbers, and pictures belong together
A quantitative display is not just geometry. Titles frame the question, labels carry units and definitions, annotations explain events or anomalies, and captions preserve context. Integrate these near the data rather than forcing viewers to decode a remote legend or footnote.
Mechanism over slogans
Do not merely say "chartjunk" or "show the data." Explain the mechanism: what the current design makes hard to compare, why that matters, and how the proposed change improves inference.