This entry is part of the Chart Index, the reference library for the Chart Design Field Guide.

The Marimekko chart — named after the Finnish textile designer Marimekko, in homage to its grid-like geometry — is a stacked bar with a twist: every bar's width is also data. Where a standard stacked bar uses uniform bar widths, the Marimekko makes width proportional to one quantitative variable and stack height proportional to another. The result is a grid of rectangles where each cell's area is a meaningful product of two variables.

It is the chart used in strategy consulting decks since the 1970s, often for market share by segment × revenue size. The form is dense, information-rich, and — for the right question — uniquely communicative. For most others, it is a way to make two simpler charts confusing simultaneously.

What it is

A Marimekko chart maps a primary categorical variable to bar position along the x-axis, a secondary categorical variable to colour-coded segments within each bar, the bar's height to the share of the segment within the bar (typically 0–100%), and the bar's width to a third quantitative variable (the bar's total size). Each cell's area thus represents segment value within its category × category size.

The hero chart above puts six industry verticals across the x-axis, each bar as wide as that industry's total cloud spend and each stacked to 100 per cent with four workload types. Read it three ways: the widths rank the industries (financial services spends more than the next two combined); the stacks compare the mix (health is storage-heavy, mining is compute-heavy); and the cell areas give absolute spend per workload per industry — the large blue block at the left is the single biggest line item in the whole market. Randomise the data and both encodings move at once: the bars change width and the stacks re-proportion, which is exactly the double reading the form asks of its audience.

When to use it

Marimekko charts are the right choice when:

  • You have a two-categorical, two-quantitative dataset where both quantitative dimensions matter.
  • The reader's question is "how big is each cell, in context of both dimensions?"
  • The form is a strategic or planning deliverable — market sizing, portfolio review.
  • You have a small enough matrix (3–8 categories × 2–5 segments) for cells to be readable.
  • The audience is familiar with the form or willing to read its legend carefully.

When not to use it

  • Most business communication. The form's complexity demands attention; for executive summaries, a stacked bar or pair of bars usually wins.
  • More than 8 categories. Bar widths shrink to ribbons; cells become unreadable.
  • More than 5 segments per bar. The stack becomes a tower of colour bands.
  • When width does not have a meaningful encoding. A Marimekko with uniform widths is just a 100% stacked bar — use that instead.

Design principles

Sort widths by size

The widest bar should be leftmost (or rightmost, consistently); progressing in decreasing width gives the chart visual rhythm and helps the eye anchor on the major categories first. Random width ordering produces a wall of irregular columns. The two figures below hold the same market; only the bar order differs.

Bars in alphabetical order
Six industries sorted A to Z. The widths lurch wide, narrow, wide again, and the reader has to hunt for the biggest market rather than being handed it.
Bars sorted by width, widest first
The same six industries in descending order of total spend. The chart now reads left to right as a ranked list of markets, and the stacks can be compared as the eye moves along a steadily narrowing staircase.

Order segments consistently

Every bar must stack its segments in the same order. The reader builds a mental template — bottom is Compute, then Storage, then Data — and applies it across bars. Re-ordering by segment value per bar destroys comparability. Only the bottom segment sits on a common baseline, so put the segment the reader most needs to compare at the bottom of the stack.

Use one colour family for segments

Stacked segments share an axis of meaning. A small, ordered set of related hues lets the reader compare segment-by-segment across bars; saturated rainbow palettes turn the chart into a quilt. The figures in this entry use two related cool hues for the infrastructure workloads, one warm hue for data, and a neutral for the remainder — four colours, one of which is deliberately quiet.

Read the area as a product

A single cell's width comes from its bar (category size), its height from the stack (segment share), and its area is the product (absolute segment value). The stacked bar below shows the same market with uniform widths; the Marimekko that follows restores them. Compare the compute segment for Media across the two: its share is the same in both, but the Marimekko shows that the absolute spend is tiny, because the bar it sits in is narrow.

As a 100 % stacked bar: shares only
Uniform bar widths. Every industry gets the same visual weight, so Media's compute segment looks as important as Financial's. The totals have been thrown away.
As a Marimekko: shares and sizes
The same shares, now with bar width set by each industry's total spend. Cell area is absolute spend. Media's compute segment is the same height as before and a fraction of the area; Financial's is the largest cell on the chart.

Label totals on the x-axis

Below each bar, place the category's total value or label. Above each bar (or alongside), the segment percentages. The chart's two encodings each need a clear annotation channel; mixing them confuses. Where the tool cannot print totals under the bars, the caption or subtitle should say what the widths measure and in what unit.

Add a thin top axis for width

The x-axis is not a category axis in the traditional sense — bar positions are proportional to widths. A short axis showing 0%, 25%, 50%, 75%, 100% along the top helps the reader read bar widths as shares of the total.

Avoid 100% Marimekko when you mean stacked bars

If your bars all have the same total width (i.e., you are normalising to 100%), you have a 100% stacked bar chart, not a Marimekko. The Marimekko's information density is in the variable widths.

Resist the urge to chart everything in one

The form's danger is that it can encode three or four dimensions in one chart. Just because you can does not mean you should. If the reading task does not need the area = product interpretation, use two simpler charts: a bar of the category totals beside the 100 % stacked bar shown earlier answers the same two questions separately, and most readers will answer them faster.

The simpler pair, part one: totals as a bar
Industry totals alone, sorted. Paired with the 100 % stacked bar above, this covers everything the Marimekko encodes — size and mix — in two charts that each ask one question. Use the pair unless the reader genuinely needs cell areas.

Anatomy

The composition of a Marimekko
An anatomical guide

A Marimekko's anatomy is variable-width stacked bars laid out with both axes encoded. The chart is dense by design; the design challenge is to ensure each cell is readable in isolation and the matrix is readable as a whole.

  • Mosaic plot — the statistician's name for the same form, used for contingency tables in categorical analysis.
  • 100% stacked bar — uniform widths, normalised heights. The Marimekko's simpler cousin.
  • Treemap — for area-encoded part-to-whole without the cross-tabulation structure.
  • Sankey diagram — for flows between two categorical dimensions.
  • Heatmap — for the same matrix structure with magnitude encoded as colour instead of area.

Reading list

  • Hartigan, J. & Kleiner, B. (1981). Mosaics for Contingency Tables. The original statistical mosaic plot.
  • Friendly, M. (1994). Mosaic Displays for Multi-Way Contingency Tables. The modern extension.
  • Tufte, E. (1990). Envisioning Information. On variable-width bar variants.