Small Multiples

Small multiples are the same chart repeated across groups in a grid — one panel per region, month, model, or segment — holding the chart type, variables, encoding, units, and axes constant. Only the subset changes.

The repetition is the mechanism. Because every panel is built identically, a reader learns to read one and applies that reading to all the rest, so attention goes to the differences between panels rather than to decoding each chart. The grid works like a table made of pictures: corresponding positions can be compared across cells.

The separation also protects the analysis. Twelve monthly panels of temperature keep variation within a month distinct from variation between months; one combined display of the same values yields a median describing no season and a spread that is mostly seasonal difference. Grouping first is what makes each distribution mean something.

Shared scales are not a refinement but a requirement. If one panel’s axis runs to 5% and another’s to 15%, equal-looking lines represent different values and genuinely different panels can look alike. Free scales are legitimate for comparing shapes rather than magnitudes, but that choice must be stated on the display.

Two further decisions carry weight. Panel order — geographic, by average level, by trend shape — reveals structure that alphabetical order hides. And a small multiple is the wrong choice for two or three series that rarely overlap, or when precise differences at the same instant matter; there one well-labeled panel is clearer. Repeated panels compare patterns across many groups; a shared panel compares a few values directly.

References:

• Cleveland & McGill, Graphical Perception (JASA 79:387, 1984)