Stacked Area Charts: Great for Composition Trends, Weak for Comparing Middle Categories

TLDR

The main stacked area chart limitations come from the moving baselines of the upper layers. A stacked area chart is useful when the primary story is the size of a total and its broad composition over time. It is much weaker when readers need to compare individual categories, especially categories in the middle of the stack. Use a regular stacked area chart for absolute amounts, a 100% stack for shares of a clearly defined whole, and a line chart, small multiples, grouped bars, dot plot, or table when precise comparisons matter.

The key is to decide whether the reader needs to understand the total, the composition, or the individual series. A stacked area chart can show the first two together, but it usually sacrifices clarity on the third. That tradeoff is not a flaw in the underlying data. It is a consequence of how the chart encodes values.

How to read a stacked area chart

In a standard stacked area chart, categories are added vertically at each point in time. The upper boundary of the complete stack represents the combined total. The value of an individual category is represented by the vertical thickness of its colored band, not by the height of its upper edge alone.

Suppose the stack contains categories A, B, and C. Category A begins at zero, so its upper boundary directly represents A. Category B begins at the top of A, meaning B must be read as the distance between two boundaries. Category C begins at A plus B, so its value is likewise the thickness between its lower and upper boundaries. The final top contour represents A plus B plus C.

This is why the chart works well for the combined total and broad rises or falls. The CDC's guidance on stacked area charts recommends the form for overall trends rather than specific values or small differences between groups.

The moving-baseline problem

Only the bottom category has a fixed zero baseline. Every category above it rests on a boundary determined by all categories below it. When those lower categories change, the baseline for an upper category moves as well.

Consider a hypothetical middle category that increases from 20 to 25 units. If the categories beneath it rise much faster, both of its boundaries will move sharply upward. A reader may notice the movement of the band but struggle to judge whether its thickness grew by five units. The visual motion reflects both the category's own change and changes in the layers below it.

The problem becomes especially important when two middle categories differ only slightly. Comparing their thicknesses requires estimating distances between separate, changing boundaries. By contrast, points or lines plotted against one shared scale can be compared directly. Cleveland and McGill's foundational research on graphical perception supports the broader principle that position along a common scale enables more accurate quantitative judgments than less direct encodings.

The bottom layer is therefore usually the easiest individual category to compare through time. The top contour is easy to follow too, but it represents the cumulative total rather than the top category. Interior layers receive neither advantage.

Why an individual category can appear to decline

A category can rise in absolute value while accounting for a smaller part of a rapidly growing total. In a stacked display, that combination may produce a band that looks flatter or proportionally thinner even though its numerical value increased. CDC guidance specifically warns that a group can increase in absolute terms while appearing to decline relative to faster-growing groups.

This distinction is common in market-share reporting. A company can sell more units while losing unit share if the whole market grows faster. Revenue share can tell a different story again if prices or product mix change. The same denominator discipline discussed in revenue share versus unit share applies to stacked charts: identify whether the chart shows amounts, percentages, revenue, units, users, or another measure before interpreting a band as growth or decline.

Standard versus 100% stacked area charts

A standard stacked area chart and a 100% stacked area chart answer different questions. Choosing between them requires deciding whether total magnitude matters.

Chart type	What the vertical scale means	What it reveals	What it obscures
Standard stacked area	Absolute values and their cumulative total	Changes in the total and broad changes in composition	Precise trends for middle and upper categories
100% stacked area	Each time point sums to 100%	How category shares change within a defined whole	Changes in the absolute total and often small share differences
Multi-series line chart	Each series uses a common quantitative scale	Individual trends and direct category comparisons	Composition is less visually prominent
Small multiples	Each category receives a separate panel, usually on aligned scales	Individual shapes and trends with less overlap	The combined total is not immediately visible

Use a standard stack for amounts and total change

A standard stack is appropriate when the total has substantive meaning and category values are additive. Examples include electricity generation by source, total sales by product group, or total spending divided among mutually exclusive categories. The reader can follow the top boundary to see whether the whole is growing or shrinking while using band thickness to understand broad compositional shifts.

The form becomes less useful if the primary question is whether one middle category grew faster than another. A line chart or small multiples will usually answer that question more directly.

Use a 100% stack for shares of a defined whole

A 100% stacked area chart rescales every time point so the categories sum to 100%. Microsoft describes percent-stacked area charts as a way to show proportional data over time, while pointing to line charts for time trends and column charts for comparisons among multiple series.

Normalization makes share changes visible but removes the size of the total. A category holding 30% could represent 30 out of 100 units in one period and 300 out of 1,000 in another. Those points occupy the same thickness in a 100% chart even though the absolute amount differs tenfold.

If total volume materially affects the story, pair the percentage chart with a total line, a second chart, or direct total labels. Otherwise, readers may interpret stable shares as stable activity when the underlying market or population changed substantially.

The denominator test for a 100% stack

Before using a 100% stacked area chart, complete this sentence: “Each vertical slice shows the percentage of _ accounted for by these categories.” If the blank cannot be filled with a clear, meaningful denominator, the chart is not ready.

• Define what enters the total and what is excluded.
• Confirm that categories use the same unit and time period.
• Check whether categories are mutually exclusive or can overlap.
• State whether the stack covers the complete population or only selected categories.
• Keep the denominator consistent, or disclose changes in coverage or definitions.
• Verify that missing or unknown cases are handled consistently across periods.

Multiple-response survey questions are a common problem. If respondents can select more than one answer, the response percentages may sum to more than 100%. Forcing those responses into a 100% stack either creates a misleading denominator or requires a different measure, such as share of all selections. Separate lines, dots, or bars are generally clearer for overlapping response rates.

An “Other” category can make a stack mathematically complete, but its composition may change over time. If several small categories move into or out of “Other,” an apparent trend may reflect classification changes rather than behavior. Keep category definitions stable and disclose material regrouping.

Stack order changes readability, not the data

Reordering a stack does not change the underlying category values or the total. It changes which series receives the stable zero baseline and which bands must be read between moving boundaries. That can substantially alter which trend looks most prominent.

The Maersk Design System advises placing a large or stable category at the bottom and smaller or more volatile categories nearer the top. It also recommends a zero baseline and cautions against using stacked areas for exact comparisons. Research by Byron and Wattenberg similarly identifies baseline choice and layer ordering as design variables connected to the legibility and appearance of stacked graphs.

Choose the order according to the reader's task rather than accepting alphabetical order automatically. Put the most important category at the bottom when its individual trend deserves the clearest baseline. A consistent logical order may be better when categories have a natural sequence, such as age bands or stages in a process. Avoid changing the order between periods, because crossing and rearranging layers make category tracking harder.

Too many categories turn composition into clutter

Every additional layer introduces another boundary, color, and legend lookup. Thin middle bands become hard to trace, while volatile layers create an irregular stack whose local movements are difficult to attribute.

Maersk presents roughly three to five categories as a practical readability guideline rather than a universal law. The right limit depends on chart size, label placement, color contrast, time resolution, and how similar the category values are. Five stable, well-separated layers may work; three thin, frequently crossing categories may still be confusing.

Combining minor categories into “Other” can reduce clutter, but only when their individual identities are not important. If readers need to compare those smaller categories, small multiples or a table preserve more information. Direct labels at the right edge can also reduce repeated legend lookups, provided the labels remain legible.

Better alternatives for precise comparisons

The best replacement depends on the question. Stacked area charts are not inherently bad; they are simply optimized for a different task than exact category comparison.

Reader's main question	Better default	Reason
How did each category change over time?	Multi-series line chart	All series share a common scale, making individual trends easier to compare.
What is the shape of each category's trend?	Small multiples	Separate aligned panels remove overlap and moving baselines.
Which category was larger on a few selected dates?	Grouped bars or dot plot	Values at discrete dates can be compared from a common baseline or scale.
How did the composition of a fixed whole change?	100% stacked area chart	The chart emphasizes proportional shares rather than total magnitude.
How did both the total and broad composition change?	Standard stacked area chart	The top contour shows the total while the bands show broad composition.
What were the exact values?	Table or direct labels	Readers do not have to estimate band thicknesses.

Grouped bars and dot plots are particularly useful when the analysis concerns only a few dates, such as the start and end of a period. A full area chart implies continuous development and can become visually dense when the actual comparison is between two or three snapshots.

Small multiples are usually the strongest alternative when many individual trends matter. Use consistent vertical scales when magnitude comparisons are important. If each panel uses a different scale, label that choice prominently: independent scales make shapes easier to inspect but can exaggerate small movements and prevent direct magnitude comparisons.

A similar comparison limitation applies to stacked bar charts. The first segment begins from a common baseline, while upper segments begin at cumulative values that vary from bar to bar. Stacked bars work for composition across discrete groups, but grouped bars, dots, or separate panels are better when upper segments must be compared precisely. This reflects the same reason that bars usually outperform pie slices for close comparisons: a shared scale makes differences easier to judge.

A practical checklist before publishing

1. Write down the primary question: total, composition, individual trends, or exact values.
2. Verify that the categories are additive and use the same unit and period.
3. Choose a standard stack for absolute amounts or a 100% stack for shares.
4. For a 100% stack, define the denominator and confirm that categories form a meaningful whole.
5. Put the category that most needs a stable baseline at the bottom, unless a natural sequence is more important.
6. Reduce the number of layers or switch to small multiples when bands become difficult to trace.
7. Use a zero baseline for a standard stacked area chart.
8. Label units, dates, category definitions, missing values, and material changes in coverage.
9. Add a companion total line, line chart, or table when readers need information the stack hides.

Negative values are another warning sign because they complicate the cumulative baseline and the meaning of band thickness. Irregular time intervals can also mislead if equal horizontal spacing implies equal periods. Missing observations should not be silently connected as though data were recorded continuously. In each case, the chart should make the data structure visible rather than relying on the filled area to smooth over gaps or incompatibilities.

Conclusion: use stacked areas for the whole, not precise parts

Stacked area charts are strongest when the message is that a meaningful total changed over time and its broad composition changed with it. Their central weakness is the moving baseline: only the bottom layer can be read directly from zero, while middle and upper categories must be estimated from band thickness.

Use a standard stack when total magnitude matters and a 100% stack when proportional composition is the real question. When the reader must compare individual categories, switch to lines, small multiples, grouped bars, dots, or a table. If both views matter, do not ask one chart to do everything—pair the stacked area with a companion chart that restores the common baseline or exact values.

References

1. Stacked Area Chart | COVE | CDC
2. Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods: Journal of the American Statistical Association: Vol 79, No 387
3. Area Charts in Power BI Paginated Reports – Power BI | Microsoft Learn
4. Stacked area chart – Maersk Design System
5. Stacked Graphs – Geometry & Aesthetics