Chris Parmer

October 24, 2026

Marimekko Charts - A Complete Guide

TL;DR

A Marimekko chart is a 100% stacked bar chart where each bar is as wide as its share of the total, so every segment's area is its share of the whole.
Marimekko charts are also known as Mekko charts, mosaic plots and percent stacked bar plots, and are similar to 100% stacked bar charts and treemaps.
Use one when the data has two categorical dimensions and both shares matter, such as market share by region and company or spending by type of care and payer.
Use grouped bars instead when readers need exact values, and fold bars thinner than about 1% of the width into an "Other" bar.
You can make a Marimekko chart with Plotly from go.Bar, with a width and offset=0 for each bar and barmode="stack", and any coding agent can build one in a minute.

What is a Marimekko chart?

Marimekko charts look like this:

Marimekko chart of 2024 electricity generation for the 12 largest generating countries and the rest of the world, split by source. China's bar is 33% of the width and 58% coal; its coal segment, 5,828 TWh or 19% of world generation, is larger than the whole US bar (4,391 TWh). France is 68% nuclear and Brazil 55% hydro

Data: Ember, Yearly Electricity Data, 2024.

How do you read a Marimekko chart? In the example above, notice these things:

Each bar is one country, and its width is that country's share of the world's electricity in 2024. China's bar takes about a third of the chart, the bars get narrower from left to right, and every other country is grouped into the wide Rest of world bar at the end.
Each bar is split by source and always reaches 100%, so the heights show a country's mix. France's bar is mostly purple nuclear, Brazil's is mostly blue hydro, and India's is mostly dark gray coal.
Width times height is area, so the size of a segment is the actual amount of electricity. That's why China's dark coal block is bigger than the whole United States bar next to it.
The sources are stacked in the same order in every bar, with the fossil fuels in grays at the bottom and nuclear and renewables above them, so you can compare how much of each bar is gray. Germany's wind segment is the tallest of any country.
Only the coal segments share a baseline at 0%, so they are easier to compare across countries than the segments floating in the middle. The narrowest bars, such as Saudi Arabia and Iran, are too thin to label, and you hover over them to read the numbers.

It's often used for charts like these:

Marimekko chart of S&P 500 market value at the end of 2025 by sector and company size. Information technology is 33% of the width and 67% of it is in companies worth over $1 trillion; communication services is 75% trillion-dollar companies; real estate, utilities and materials have no company over $200 billion

A market map. Ten companies worth over $1 trillion each were 41% of the S&P 500's $62.3 trillion at the end of 2025, and 67% of the technology sector. Data: SEC EDGAR (shares) and Yahoo Finance (prices), index members as of October 2026.

Marimekko chart of 2024 Medicare Part D spending for the 10 manufacturers with the most spending, each bar split into its top drug, second and third drugs and all others. Eliquis is 71% of Bristol Myers Squibb's $29.1 billion and Ozempic 70% of Novo Nordisk's; Humira is 26% of AbbVie's

Revenue by company and product. Eliquis was 71% of Bristol Myers Squibb's $29.1 billion of Medicare Part D spending in 2024, while Humira was 26% of AbbVie's. Data: CMS, Medicare Part D Spending by Drug, 2024.

Mosaic plot of 1,309 Titanic passengers by class and sex, with survivors in blue. 97% of first-class women survived and 15% of second- and third-class men; third-class men are the widest bar with 493 people

A contingency table of counts. 97% of first-class women on the Titanic survived, against about 15% of men in second and third class, and third-class men were the largest group at 493 of 1,309 passengers. Data: titanic3, Encyclopedia Titanica via Vanderbilt Biostatistics.

A short history

Statisticians call this chart a mosaic plot. John Hartigan and Beat Kleiner introduced it in 1981 to show contingency tables (Mosaics for contingency tables), and Michael Friendly extended it to tables with many variables in 1994 (Mosaic displays for multi-way contingency tables), using a table of hair and eye color of 592 students that became the standard example:

Mosaic plot redrawn with Plotly of the hair and eye color of 592 students (Snee, 1974). Brown hair is the widest bar with 286 students; 74% of the 127 blond-haired students had blue eyes and 63% of the 108 black-haired students had brown eyes

In business strategy work the same chart is used as a "market map" and is usually called a Marimekko or mekko chart, a name commonly attributed to its resemblance to the prints of the Finnish design company Marimekko. Friendly traces the idea back to earlier area-proportional displays in A brief history of the mosaic display (2002).

Alternatives to a Marimekko chart

Here's the same table of US health care spending drawn six ways. Some are better for the question, some worse, some just different:

The same 2024 US health spending by type of care and payer drawn six ways: a Marimekko chart, 100% stacked bars with equal widths, stacked bars in dollars, grouped bars in dollars, a treemap of type of care and payer, and a heatmap of each payer's share of each type of care

Grouped bars (better for exact amounts). Every bar sits on a shared baseline, so you can see that private insurance paid $559 billion of hospital care and $505 billion of physician care, which a Marimekko only shows as areas. You lose the part-to-whole.
100% stacked bars (worse when sizes differ). The mix of each type of care is the same as in the Marimekko, but hospital care ($1,635 billion) gets the same width as dental care ($374 billion), so the size of each one is lost.
Stacked bars in dollars (works too, for totals). The heights show the totals, but the mixes of the small bars become hard to read and compare.
Treemap (different). Shows sizes and nesting with more room for labels, but its rectangles share no baseline, so mixes are hard to compare across groups. See our treemap guide.
Heatmap of shares (better for comparing mixes). Each cell is a share, so it compares every payer across every type of care exactly (out of pocket is 32% of other care and 2.5% of hospital care), but it drops the size of each type. See our heatmap guide.

How to make a Marimekko chart

With AI

ChatGPT, Claude Code, GitHub Copilot, Cursor and other AI tools are excellent at writing Plotly code. Name Plotly in the prompt, name the construction (because "marimekko" alone sometimes gets you a treemap), say what width and height mean, and list the checks you want:

Make a plotly Marimekko chart from sales.csv (columns: region, product, revenue_usd).

- One bar per region, width = region's share of total revenue; x = running sum of widths, offset=0.
- Segments = products, height = share of the region's revenue; one go.Bar trace per product,
  barmode="stack", bargap=0, a 0.5% gap between bars.
- Sort regions by revenue; fold regions under 1% into "Other regions", drawn last.
- Same product order and color in every bar.
- Label each bar under the axis with name and share; label segments only where they fit.
- Hover: region, product, revenue and share.
- A title that states the takeaway and a subtitle that says what width and height mean.
- Assert the table sums to the total before plotting.

With Plotly Studio

Plotly Studio is a desktop app where you point an AI at a dataset and it builds charts and a Dash app around your questions. Ask for a Marimekko chart and say what the width and the height should mean:

Show a Marimekko chart of electricity generation with one bar per country, as wide as its share of world generation, split into sources by their share of that country's generation. Use one fixed color per source, sort the countries by size, and add a dropdown to switch the year.

Make a Marimekko chart in Python or JavaScript

This is the code AI tools write under the hood. In Python, this helper takes a long table with one row per pair, computes each bar's width and left edge, and stacks one go.Bar per sub-category. Here it is on the Titanic's 1,309 passengers by class:

import pandas as pd
import plotly.graph_objects as go


def make_marimekko(df, x, color, value, gap=0.005):
    t = df.pivot_table(index=x, columns=color, values=value, aggfunc="sum", fill_value=0)
    t = t.loc[t.sum(axis=1).sort_values(ascending=False).index]  # widest bar first
    widths = t.sum(axis=1) / t.values.sum()   # bar width = share of the grand total
    lefts = widths.cumsum() - widths          # left edge of each bar
    shares = t.div(t.sum(axis=1), axis=0)     # segment height = share within the bar
    fig = go.Figure()
    for c in t.columns:
        fig.add_bar(name=c, x=lefts + gap / 2, width=widths - gap, offset=0, y=shares[c],
                    text=[f"{s:.0%}" for s in shares[c]], customdata=list(zip(t.index, t[c])),
                    hovertemplate=f"%{{customdata[0]}}, {c}: %{{customdata[1]:,}} (%{{y:.0%}})<extra></extra>")
    fig.update_layout(barmode="stack", bargap=0, uniformtext=dict(minsize=10, mode="hide"))
    fig.update_xaxes(range=[0, 1], tickvals=lefts + widths / 2,
                     ticktext=[f"{i}<br>{w:.0%}" for i, w in widths.items()])
    fig.update_yaxes(range=[0, 1], tickformat=".0%")
    return fig


df = pd.DataFrame({
    "pclass": ["1st class"] * 2 + ["2nd class"] * 2 + ["3rd class"] * 2,
    "survived": ["Survived", "Died"] * 3,
    "passengers": [200, 123, 119, 158, 181, 528],
})
fig = make_marimekko(df, x="pclass", color="survived", value="passengers")
fig.show()

In JavaScript, plotly.js builds the same chart from bar traces with a width and an offset for each bar:

<div id="chart"></div>
<script src="https://cdn.plot.ly/plotly-4.1.1.min.js"></script>
<script>
  const classes = ["3rd class", "1st class", "2nd class"];
  const counts = {Survived: [181, 200, 119], Died: [528, 123, 158]};
  const totals = classes.map((_, i) => counts.Survived[i] + counts.Died[i]);
  const all = totals.reduce((a, b) => a + b);
  const widths = totals.map(n => n / all);                    // bar width = share of all passengers
  const lefts = widths.map((_, i) => widths.slice(0, i).reduce((a, b) => a + b, 0));
  const traces = Object.keys(counts).map(name => ({
    type: "bar", name, x: lefts, width: widths.map(w => w - 0.005), offset: 0,
    y: counts[name].map((n, i) => n / totals[i]),             // height = share within the bar
    texttemplate: "%{y:.0%}"
  }));
  Plotly.newPlot("chart", traces, {
    barmode: "stack", bargap: 0,
    xaxis: {range: [0, 1], tickvals: lefts.map((l, i) => l + widths[i] / 2), ticktext: classes},
    yaxis: {range: [0, 1], tickformat: ".0%"}
  });
</script>

Without offset: 0, Plotly centers each bar on its x value and the bars overlap, so set it on every trace.

Share it

Every Plotly chart has a Share chart button (the cloud icon in its toolbar) that uploads it to Plotly Cloud and gives you a link. Only you can see it until you change its visibility.

For a live app, put the chart in a Dash app and publish it with the Plotly Cloud CLI. Coding agents can run these commands for you:

pip install "dash[cloud]"
plotly user login
plotly app publish

If your data has to stay in your own cloud or data center, publish to Dash Enterprise instead.

Dos and don'ts

Most of the work in a Marimekko chart is in the order of the bars and segments, the gaps and the colors. Here is the same health spending table both ways.

Don't sort the bars alphabetically, drop the gaps (so segments of the same color run together across bars), force raw numbers with no unit onto every segment, or leave out the width shares that are the point of the chart:

A Marimekko chart of 2024 US health spending by type of care and payer with real mistakes: bars in alphabetical order with no gaps, so the blue and orange segments of neighboring bars merge, raw numbers such as 558.9 with no unit in a serif font on every segment including an unreadable 40.6 on a thin strip, and bar labels without each bar's share of the total

Do use the same colors with the bars sorted by width, small gaps, each bar's share under it, and shares on the segments only where they fit:

The same Marimekko chart designed on purpose: bars sorted from hospital care (36% of spending) to dental and other professional care (8.3%), small gaps between bars, private insurance blue at the bottom and out of pocket orange at the top of every bar, and percentage labels such as 27% for Medicare's share of hospital care

Use color for one story

Width and height already carry two numbers, so color decides what the reader sees first. Here are four options on the same data.

One color per payer. A good default when the whole mix is the story.

The health spending Marimekko chart with one color per payer and a legend: private insurance blue, Medicare purple, Medicaid green, other gray and out of pocket orange

Highlight one payer. Out of pocket is the only color and everything else is gray, so the chart is about what patients pay.

The health spending Marimekko chart with out-of-pocket spending in orange and every other payer in shades of gray. Out of pocket is 32% of other care and products, 31% of dental and other professional care and 20% of nursing and home health

Sort by the highlighted share. Ordering the bars by out-of-pocket share turns the chart into a ranking, from 32% of other care to 2.5% of hospital care.

The highlighted health spending Marimekko chart with the bars sorted by out-of-pocket share: other care and products 32%, dental and other professional 31%, nursing and home health 20%, drugs 12%, physician and clinical 8.5% and hospital care with a thin orange strip

Label the segments directly. Writing the payer names on the widest bar replaces the legend, so readers don't look back and forth.

The health spending Marimekko chart without a legend: the hospital bar is labeled Private insurance 34%, Medicare 27%, Medicaid 20% and Other 17%, and the other care bar is labeled Out of pocket 32%

Checklist

Write a title with the takeaway and a number, and say in the subtitle what width and height mean.
Sort the bars by width, or keep a natural order (phases, age bands, wealth groups).
Keep one segment order and one color per category in every bar, with the segment the story is about at the bottom or the top.
Leave small gaps between bars (about 0.5% of the width) instead of borders.
Label the bars under the axis, turning narrow labels vertical, and label segments only where they fit.
Fold the tail of small bars into one "Other" bar at the end.

Don't

Sort the bars alphabetically or let the segment order change from bar to bar.
Use a rainbow for unordered categories, or more than about eight segment colors.
Let bars get thinner than about 1% of the width.
Print absolute values on segments without the shares.
Rely on hover in a static image.

Options to customize

These are the go.Bar and layout options that build a Marimekko chart. The full list is in the Plotly bar chart documentation.

x: each bar's left edge, the running sum of the widths before it. Use with offset=0.
width: each bar's width in x units (its share of the total). Subtract a small gap.
offset: shift of the bar from x; 0 starts the bar at x instead of centering it.
y: segment height, the share within the bar.
layout.barmode: "stack" piles the traces into one bar.
layout.barnorm: "fraction" or "percent" normalizes each stack, when you pass raw values instead of shares.
layout.bargap: set to 0 and control the gaps through width.
marker.color: one fixed color per sub-category, or grays plus one highlight.
text, texttemplate: segment labels, usually the share.
insidetextanchor, textangle: center the labels and keep them horizontal.
layout.uniformtext: a minimum font size; mode="hide" drops labels that don't fit.
customdata, hovertemplate: absolute values and both shares in the hover.
xaxis.tickvals, ticktext: bar names and width shares at the bar centers.
layout.annotations: free text, such as vertical labels for bars under about 4% wide.
layout.legend.traceorder: "normal" keeps the legend in the segment order.
xaxis.range, yaxis.range, tickformat: [0, 1] and ".0%".

More Marimekko chart examples

Ten more Marimekko charts built from real public data, each with its source, the technique that makes it work and a prompt to rebuild it.

1. US energy by sector and fuel

Marimekko chart of US primary energy consumption in 2025 by sector and fuel. Electric power is 35% of the width and the only sector with nuclear (24%); transportation is 29% and 89% petroleum; residential and commercial are mostly natural gas

The US used 96.4 quadrillion Btu of primary energy in 2025. Transportation is 29% of it and runs 89% on petroleum, while the electric power sector, the widest bar at 35%, is the only one that uses nuclear (24% of its energy). Data: US EIA, Monthly Energy Review, tables 2.2 to 2.6.

Tip: keep the sectors in a meaningful order and count electricity once, where it's generated, so the widths add up to primary energy.

Get the data from the US EIA Monthly Energy Review, tables 2.2-2.6 (eia.gov/totalenergy), annual 2025,
primary energy consumed by sector and fuel, and shape it into a table
(columns: year, sector, fuel, tbtu; sectors residential, commercial, industrial, transportation,
electric power; fuels coal, natural gas, petroleum, nuclear, renewables).

Build a plotly Marimekko chart with sectors as bars and fuels as segments.
- Petroleum orange, gas and coal grays, nuclear purple, renewables green.
- Title with transportation's petroleum share; subtitle says electricity is counted where generated.

2. Fossil CO2 by country and fuel

Marimekko chart of 2024 fossil CO2 emissions by country and fuel. China is 32% of the width and 72% coal, so its coal segment is 23% of the world's fossil CO2; the US bar is 13% wide and mostly oil and gas; Iran is mostly gas and Saudi Arabia mostly oil

The world emitted 38.6 billion tonnes of fossil CO2 in 2024. China is 32% of the width and 72% of its emissions come from coal, so coal burned in China is 23% of the global total, written on the segment itself. Data: Global Carbon Project via Our World in Data, 2024.

Tip: when one segment is the story, write its share of the whole chart (width times height) on it.

Get the data from the Global Carbon Project via Our World in Data's CO2 dataset (github.com/owid/co2-data), 2024
and shape it into a table
(columns: year, country, fuel, mt; fuels coal, oil, gas, cement and other; the 10 largest emitters plus
"Rest of world" = World minus them, including international aviation and shipping).

Build a plotly Marimekko chart: countries as bars, fuels as segments, coal near-black at the bottom.
- Annotate China's coal segment with its share of the world total.

3. US electricity by state and source

Marimekko chart of US electricity generation in 2025 for the 15 largest generating states and the other 35 states and DC as one bar. Texas is 13% of the width with 22% wind; California is 37% solar, Illinois 50% nuclear and Florida 72% gas

Texas made 13% of US electricity in 2025 and got 22% of its power from wind. The mixes differ a lot: California is 37% solar, Illinois 50% nuclear, Florida 72% gas. Data: Ember, US Electricity Data (compiled from EIA), 2025.

Tip: give the 15 largest states their own bars and fold the rest into one bar at the end, so nothing is too thin to label.

Get the data from Ember's US Electricity Data (ember-energy.org), yearly 2025, generation by state and fuel,
and shape it into a table (columns: year, state, source, twh; 50 states and DC).

Build a plotly Marimekko chart of states and sources, using the same source colors as a world chart.
- Keep the 15 largest states; fold the others into "Other 35 states and DC", drawn last.
- Set slightly negative net generation to zero and say so in the subtitle.

4. What the 12 largest US banks hold

Marimekko chart of the assets of the 12 largest US banks on June 30, 2026, by type. JPMorgan Chase is 27% of the width and 38% loans with 17% trading assets; Capital One is 66% loans; BNY and State Street, the custody banks, are mostly cash and securities, with loans at 11% and 13%

The 12 largest US banks held $15.2 trillion on June 30, 2026, 57% of all FDIC-insured bank assets. Loans are 66% of Capital One's assets and 11% of BNY's, because custody banks hold cash and securities instead. Data: FDIC BankFind Suite (Call Reports), June 30, 2026.

Tip: put the defining asset, loans, at the bottom so every bank's loan share starts at zero.

Get the data from the FDIC BankFind Suite API (banks.data.fdic.gov), latest quarter, fields ASSET, LNLSNET, SC,
CHBAL, FREPO, TRADE, and shape it into a table
(columns: report_date, bank, asset_type, usd_bn; 12 largest banks; other assets = ASSET minus the rest).

Build a plotly Marimekko chart of banks and asset types, loans blue at the bottom, other assets gray on top.

5. What US households own, by wealth group

Marimekko chart of US household assets in the second quarter of 2026 by net worth group and asset type. The top 0.1% and the rest of the top 1% hold 14% and 16% of assets, mostly stocks; the bottom 50% hold 5% of assets, 47% of it real estate and 21% consumer durables

US households held $206 trillion of assets in mid-2026. The top 1% own 51% of the stocks and mutual funds households hold, while the bottom half of households holds 5% of all assets, mostly homes and cars. Data: Federal Reserve, Distributional Financial Accounts, 2026 Q2.

Tip: keep the groups in their natural order, richest to poorest, rather than sorting by width.

Get the data from the Federal Reserve's Distributional Financial Accounts (federalreserve.gov/releases/z1/dataviz/dfa),
levels by net worth group, latest quarter, and shape it into a table
(columns: date, group, asset_type, usd_bn; groups top 0.1%, rest of top 1%, next 9%, next 40%, bottom 50%).

Build a plotly Marimekko chart with groups in that order and stocks blue at the bottom of each bar.
- Title with the top 1%'s share of household stocks.

6. What the US imports from its largest suppliers

Marimekko chart of US goods imports in 2025 by partner and product group. Mexico is 15% of the width, Canada 11% with 31% fuels, China 9.3%; Ireland is a narrow bar that is 83% chemicals and pharma, and Taiwan 86% machinery and electronics

The US imported $3.50 trillion of goods in 2025. Ireland is a narrow bar that is 83% pharma and chemicals, Taiwan is 86% machinery and electronics, and 31% of imports from Canada are fuels. Data: US Census Bureau via UN Comtrade, 2025.

Tip: give the two product groups in the story strong colors and the rest quieter ones.

Get the data from UN Comtrade (comtradeplus.un.org), US imports in 2025 by partner and HS chapter,
and shape it into a table
(columns: year, partner, product_group, usd_bn; 10 largest partners plus "Rest of world"; groups from HS chapters).

Build a plotly Marimekko chart of partners and product groups, Rest of world last.

Another way to see it: a treemap of partner and product shows the same dollars as nested areas, with Ireland's $111 billion of pharma and chemicals as one tile, but the mixes no longer line up for comparison.

Treemap of US goods imports in 2025 by partner and product group, with Rest of world ($1,160B), Mexico ($539B), Canada ($392B) and China ($327B) as the largest blocks and Ireland's chemicals and pharma tile at $111B

7. Where the world's workers work

Marimekko chart of employed people in the 12 countries with the most workers in 2025 by sector. China and India are 32% and 26% of the width; India's agriculture segment, 246 million farm workers, is 42% of its bar and larger in area than the whole US bar of 167 million

India has 246 million farm workers, more than the 167 million people working in the US. Agriculture is 42% of India's workforce and 1.5% of America's. Data: ILO modeled estimates via the World Bank, World Development Indicators, 2025.

Tip: replace one segment's percentage with its head count when the absolute number is the story.

Get the data from the World Bank's World Development Indicators (ILO modeled estimates: SL.TLF.TOTL.IN,
SL.UEM.TOTL.ZS, SL.AGR.EMPL.ZS, SL.IND.EMPL.ZS, SL.SRV.EMPL.ZS), 2025, and shape it into a table
(columns: year, country, sector, workers_mn; employed = labor force x (1 - unemployment rate)).

Build a plotly Marimekko chart of the 12 countries with the most workers by sector.

8. Who sponsors clinical trials at each phase

Marimekko chart of 98,988 interventional clinical trials started 2015-2024 by phase and lead sponsor. Phase 2 is the widest bar at 29%; industry leads 66% of phase 1 and 52% of phase 3 trials but 13% of phase 4 and 11% of early phase 1, where universities and hospitals lead

Of 98,988 trials started in 2015-2024 with a phase, industry leads 66% of phase 1 and 52% of phase 3 trials but only 13% of phase 4. Universities and hospitals lead the rest. Data: ClinicalTrials.gov API, interventional trials.

Tip: keep the phases in pipeline order, not sorted by width.

Get the data from the ClinicalTrials.gov API v2 (count queries), interventional trials started 2015-2024,
and shape it into a table
(columns: phase, sponsor, trials; phases early 1, 1, 1/2, 2, 2/3, 3, 4 made exclusive; sponsor from LeadSponsorClass).

Build a plotly Marimekko chart with phases in pipeline order and industry blue at the bottom.

9. At what age people die in each WHO region

Marimekko chart of deaths in 2023 by WHO region and age. Western Pacific is 29% of the width with 63% of deaths at 70 and over; Africa is 14% wide and 33% of its deaths are children under 5, shown in red; Europe's under-5 share is 1%

A third of deaths in Africa in 2023 were children under 5, against 1% in Europe, where 69% of people died at 70 or older. Africa had 56% of the world's under-5 deaths. Data: WHO, Global Health Estimates 2023.

Tip: one red highlight and every other band in grays gives the chart a single story.

Get the data from the WHO Global Health Estimates (xmart-api-public.who.int, GHE_REGIONS), 2023, all causes,
and shape it into a table (columns: year, region, age_group, deaths; under 5, 5-14, 15-49, 50-69, 70+).

Build a plotly Marimekko chart of WHO regions and age groups, under 5 red at the bottom, others gray.

Another way to see it: a bar chart of under-5 deaths in each region gives the counts directly, 2.81 million in Africa, which the Marimekko only shows as an area.

Bar chart of deaths of children under 5 in 2023 by WHO region: Africa 2.81 million, Eastern Mediterranean 844 thousand, South-East Asia 821 thousand, Western Pacific 253 thousand, Americas 174 thousand and Europe 83 thousand

10. The world's population by continent and age

Marimekko chart of the 2023 world population by continent and age. Asia is 59% of the width; Africa is 18% and 39% of its people are under 15; Europe's over-65s are 20% of its bar; Oceania is a 0.6% sliver

Africa has 18% of the world's 8.09 billion people and 29% of its children under 15. In Europe, people over 65 (150 million) outnumber children under 15 (116 million). Data: UN World Population Prospects 2024 via Our World in Data.

Tip: a 0.6% bar like Oceania's is fine to keep when it's labeled vertically, because it shows how small it is.

Get the data from the UN World Population Prospects 2024 via Our World in Data (population-by-age-group), 2023,
and shape it into a table (columns: year, continent, age_group, people).

Build a plotly Marimekko chart of continents and age groups, children warm colors at the bottom,
over-65s purple on top.