Edward Tufte called them "the best design solution for a wide range of problems." Small multiples — a series of charts with the same structure, each showing a different slice of the data — are the most underused technique in data visualization.
The concept is simple. Instead of layering six time series on one chart (producing an unreadable tangle of lines), show six small charts side by side, each containing one series. Same axes, same scale, same visual grammar. The reader's eye moves across the grid and comparisons emerge naturally.
Why They Work
Small multiples exploit a specific cognitive mechanism: once the reader learns to decode the first panel, every subsequent panel is decoded instantly. The learning cost is paid once. The information payoff multiplies with each additional panel.
This is fundamentally different from a complex single chart. A single chart with six overlapping series forces the reader to disentangle colors, trace intersecting lines, and constantly reference the legend. The cognitive load increases with each additional series. Small multiples reverse this relationship.
Consider a dashboard showing monthly revenue by product line. The typical approach layers all lines on one time series chart. With three products, this works. With eight, it is unreadable. Small multiples — eight small charts in a 2x4 grid — make the pattern in each product line immediately visible while preserving the ability to compare across the grid.
Design Requirements
For small multiples to work, several conditions must hold:
Consistent scales. Every panel must share the same x-axis and y-axis range. If one panel uses a different scale, the entire grid becomes a lie. The reader's eye compares positions across panels, and those comparisons must be mathematically valid.
Minimal decoration. Each panel should contain the minimum ink necessary. Axis labels can appear on the outer edges of the grid only. Grid lines, if used at all, should be faint. The data should dominate.
Clear labels. Each panel needs a clear, short title. This is often a category name — a country, a product, a demographic group. The title should be positioned consistently and should not compete with the data.
Logical ordering. The sequence of panels matters. Alphabetical ordering is the default but rarely the best choice. Ordering by the variable of interest — the panel with the highest value first, or the most change, or the most recent anomaly — turns the grid into a ranked display.
Where They Excel
Geographic comparisons. Showing the same metric across 20 countries is impossible in a single chart. A 4x5 grid of small line charts, each showing one country, makes every trend visible and every comparison possible.
Time-of-day patterns. A 7-column grid showing web traffic by day of week, with each column containing a 24-hour profile, reveals daily rhythms that a single overlay chart would obscure.
A/B test results across segments. Rather than showing aggregate results, small multiples can display the effect of a test across user segments simultaneously. The panels where the treatment failed stand out immediately.
The Resistance
Despite their effectiveness, small multiples face resistance in practice. Executives often want "one chart that tells the whole story." Product managers ask for consolidated views. The instinct to compress everything into a single visualization is strong.
This instinct is wrong. A single chart that tries to show everything shows nothing clearly. Small multiples trade visual density for perceptual clarity. The grid takes up more space, but every square centimeter of that space communicates.
Mike Bostock's work with D3.js popularized small multiples on the web, demonstrating that the technique works as well on screens as it does in print. Observable notebooks use small multiples extensively, and the results speak for themselves: complex datasets rendered comprehensible through repetition and structure.
The technique is centuries old. It predates computing entirely. And it remains, as Tufte wrote, among the best tools available.