Stacking suits totals, not per-series trends

What it is

A conventional nonnegative stack treats two questions asymmetrically. The outer boundary's position above a fixed zero line encodes the total, while an internal layer's thickness encodes a component. Stacking can therefore support questions such as when the whole grows, where it peaks, and how composition changes broadly, yet is usually weak for comparing subtle trends across components. “Not suited” is a task-fit judgment, not a prohibition: few series, large changes, labels, or interactive focus may make a component task workable.

Why it happens

The total is the sum of all layers, and the silhouette directly displays that derived quantity. An internal boundary combines the target with the cumulative values below it. Its rise or fall does not reveal the target's rise or fall; readers must instead compare the distance between two boundaries. Area fill supports layer identity and continuity, while the line-like silhouette supports the sum, but neither restores a common baseline to every component. More layers, crowded boundaries, interpolation across missing data, or volatile lower series further increase the cost of separating component trends from the whole.

Studying it

Start by separating workflow questions about totals, components, shares, rankings, and anomalies. Compare absolute stacks, standalone lines, small multiples, and interactive highlighting while holding data, horizontal domain, scale range, and available labels comparable. Manipulate layer count, target position, trend amplitude, correlation, and missingness. Measure trend direction, turning-point localization, and magnitude estimates, and note whether readers follow the cumulative edge by mistake. Report total and component tasks separately; one chart-preference score would conceal the task interaction.

Where it stops holding

A total is meaningful only when components have compatible units, definitions, and additive semantics. Stacking unlike measures does not create an interpretable whole. Positive and negative values may cancel, so one silhouette no longer represents intuitive activity; separate positive and negative totals or choose another view. Missing is not zero, and filled interpolation can invent a trend. A streamgraph's moving baseline gives up zero-based total reading to emphasize temporal distribution and flow among many nonnegative series. Uncertainty in a sum may also depend on covariance and cannot always be formed by adding component intervals.

Applying it

• State the total question served by the silhouette. If the primary question is how one component changes, use a standalone line, small multiples, or a selection that rebases the target.
• When both levels matter, pair a total overview with linked component views and keep filters, time windows, and units synchronized.
• Mark missing and interpolated periods and state the aggregation definition. Use an intelligible diverging treatment for negatives rather than hiding activity through cancellation.
• Give screen readers structured navigation through the total, component series, and key turning points. Validate with separate total and component questions, not a generic comprehension prompt.

Related

• Same group: U2.04.1 Only the bottom series of a stacked chart owns the common baseline · U2.04.2 Upper series read only by band thickness, at markedly lower precision · U2.04.4 Percent stacking discards the totals · U2.04.5 Stack order decides which series stay readable
• Nearby: U2.17.1 A mismatched chart forces readers to compute the answer mentally · U2.15.1 Small multiples replace overlay with repeated like-for-like panels
• Search terms: stacked area · total trend · component trend · task matching