What if the size of a province reflected what it contributes, rather than the land it occupies?
For Day 15 of the 30 Day Map Challenge, the theme is Inside Out. After Day 14's attempt to map everything inside Skukuza Rest Camp, I wanted to do the opposite: take a familiar map of South Africa and deliberately distort it.
On a conventional map, the Northern Cape dominates. It covers 30.5% of South Africa's land area. Gauteng, meanwhile, occupies just 1.5%, barely visible beside its neighbours. But Gauteng contains 25.5% of the country's population, while the Northern Cape has only 2.2%.
So which province should look bigger?
The answer depends on what the map is trying to communicate. A geographical map shows where things are and how much space they occupy. A cartogram uses that space to represent something else.
For this challenge, I built four cartograms that resize South Africa's nine provinces according to their share of the national population or economic output in 2025. The country stays the same, but the shape of its provinces changes completely.
Explore the interactive report
The report places two maps side by side.
On the left is South Africa's familiar provincial geography, with provinces sized according to their actual land area. On the right is a contiguous cartogram, where each province is resized according to the selected measure.
Use the metric selector to switch between:
- Population: Each province's share of South Africa's estimated 2025 population.
- Agriculture: Its share of national agriculture, forestry and fishing output.
- Mining: Its share of mining and quarrying output.
- Finance: Its share of finance, real estate and business services output.
The province colours remain recognisable across the two representations, making it easier to follow a province as its outline changes.
Beside the maps, the comparison table shows each province's share of land area, its share of the selected measure, and the difference in percentage points. Two cards identify the largest expansion and contraction for the selected metric.
The difference is measured against land area, not against a previous year. A province expanding on the cartogram doesn't necessarily mean its economy has grown. It means its share of the selected measure exceeds its share of South Africa's land.
The report also doesn't show where activity is concentrated within a province. Every province is treated as one statistical unit.
Interactive report: Day 15, Inside Out
Starting with South Africa's provincial boundaries
The boundaries come from the South Africa Subnational Administrative Boundaries dataset, distributed through the Humanitarian Data Exchange (HDX) as part of the United Nations' Common Operational Datasets.
I used the administrative level 1 GeoJSON, containing South Africa's nine provinces. Each province has a name, an administrative code and a polygon defining its boundary. The administrative codes, ZA1 to ZA9, became the common identifiers linking the boundaries to the statistical data. The original GeoJSON is used directly in the left-hand map.
It also provides the geographical starting point for creating the distorted versions.
Getting the population and economic data into the same year
My first attempt used Census 2022 population figures alongside provincial economic data for 2023.
That would have worked as a cartographic demonstration, but it introduced an awkward comparison: the maps represented different years.
South Africa's population has changed since the census, and some of those changes are large enough to affect the cartogram. Gauteng's share, for example, increases from approximately 24.3% in Census 2022 to 25.5% in the 2025 estimates.
So I rebuilt the data using 2025 throughout.
Population: 63.1 million people
Population comes from Statistics South Africa's Mid-year Population Estimates 2025. These are estimates rather than a new census count. The national total is approximately 63.1 million, with provincial populations ranging from about 1.38 million in the Northern Cape to 16.10 million in Gauteng.
For the cartogram, the relevant number is each province's proportion of that national total.
Economic output: three very different industries
The economic figures come from Stats SA's Provincial Gross Domestic Product release, published in October 2026 and reporting provincial economic activity for 2025.
I selected three industries because they have different geographical distributions:
- Agriculture, forestry and fishing
- Mining and quarrying
- Finance, real estate and business services
The data is expressed in millions of rand at current prices.
National output across the nine provinces is approximately R227 billion for agriculture, R551 billion for mining and R1.77 trillion for finance. These values aren't being compared with one another. Each cartogram shows how its own industry's national total is distributed between the provinces.
That distinction matters. A province representing 20% of agricultural output doesn't necessarily produce more economic value than one representing 10% of finance.
Turning geographical boundaries inside out
The cartograms were generated using go-cart.io, a web application for producing contiguous cartograms.
A conventional choropleth map changes the colours of regions according to a value. A cartogram goes further: it changes the area of the regions themselves.
If a province accounts for 25% of the national population, the algorithm attempts to give it 25% of the cartogram's total area. It also tries to preserve the relationships between neighbouring provinces, so the map remains recognisable.
The method is based on the fast flow-based density-equalising cartogram algorithm described by Gastner, Seguy and More (2018).
Four separate cartograms
For each measure, I supplied go-cart.io with the same provincial boundaries and a table of nine values. Each cartogram was generated independently and exported as GeoJSON containing the distorted provincial polygons.
The exports preserve the province identifiers, which makes it possible to join the resulting shapes back to the original data. The difficult part wasn't generating the cartograms. It was getting four different sets of distorted geometry into one Power BI report while keeping the proportions meaningful.
From go-cart.io to WKT
The coordinates exported by go-cart.io are not geographical longitude and latitude. They're coordinates in the cartogram's own drawing space.
Icon Map Pro can render polygon geometry stored as Well-Known Text (WKT), so I converted the exported polygons to WKT and applied a uniform coordinate transformation to place them in a displayable range. This is an important distinction: the cartogram coordinates aren't real locations. They shouldn't be interpreted as latitude and longitude, and there's no meaningful distance scale.
The four transformed cartograms were combined into one table containing 36 records: nine provinces for each of four measures. Each record holds its province identifier, metric, value, actual land area, share of national land, share of the selected national total, and WKT polygon.
The original Geographic map remains separate, using the HDX GeoJSON boundaries.
Keeping the original shapes
Some of the exported polygons are surprisingly detailed, particularly where provinces become narrow or develop long, distorted edges. I chose not to simplify them further. That made the source WKT strings longer, and required care in preparing the Power BI model, but it preserved the geometry generated by go-cart.io.
It also avoided introducing a second round of shape changes simply to reduce the file size.
Checking the areas
A cartogram is only useful if its areas actually reflect the input values. After conversion, I calculated the combined polygon area for each of the four cartograms in their artificial coordinate space.
All four had effectively the same total area, with differences of less than 0.03%. Their individual provincial area shares were also close to the statistical shares used to generate them. This matters because otherwise a province could appear larger simply because the whole cartogram had been scaled differently. The maps have different outer outlines, and the geographic reference uses a different coordinate system, so the displayed zoom numbers aren't themselves a meaningful comparison.
Building the report in Icon Map Pro
The final report uses two Icon Map Pro visuals in Power BI.
The left visual displays the original provincial boundaries. The right displays the WKT polygons from the selected cartogram. I kept the layout deliberately simple: two large maps, a metric selector, a comparison table, two KPI cards and a short explanation.
A fixed reference beside a changing map
The left map stays geographical while the right map changes. I initially considered using the same zoom level for both maps, but the geometries have different coordinate spaces and outlines. Matching the numeric zoom value doesn't necessarily produce a more meaningful comparison. Instead, I framed each map so the whole country is visible and the provinces can be read comfortably.
The important comparison is the proportion of area assigned to each province, not the zoom number shown by the visual.
A table that explains the distortion
For each province, the report calculates:
Difference (percentage points) = Selected metric share − Land area share
A positive value indicates that the province occupies a greater proportion of the cartogram than it does of South Africa's actual land. A negative value indicates that it occupies less.
The table is sorted by the selected metric share, so the leading provinces rise to the top when the metric changes. This is more informative than sorting alphabetically, because it immediately reveals where the selected activity is concentrated.
Two cards for the extremes
The final addition was a pair of dynamic KPI cards:
- Largest expansion: The province with the greatest positive percentage-point difference.
- Largest contraction: The province with the greatest negative difference.
Each card shows the province, the difference and the movement from its land-area share to its selected statistical share. The measures respond to the metric slicer.
These cards turned out to be particularly useful. The table contains all nine comparisons, but the cards identify the two that explain the map's most striking changes.
What the map reveals
Population: Gauteng becomes the centre of the country
Gauteng occupies only 1.5% of South Africa's land but contains approximately 25.5% of its population. In the population cartogram, it becomes one of the largest features on the map. KwaZulu-Natal also expands considerably: 7.7% of the land, but 19.4% of the population. The Northern Cape goes the other way. Its 30.5% geographical share becomes just 2.2% of the population cartogram.
The contrast is dramatic, and it explains why maps based purely on physical geography can give a misleading impression of where people are concentrated.
Agriculture: the Western Cape expands
The Agriculture cartogram produces one of the most interesting transformations. The Western Cape accounts for 10.6% of the country's land but approximately 23.0% of agricultural output, a difference of +12.3 percentage points. KwaZulu-Natal also expands, from 7.7% of land to 15.5% of agricultural output. The Free State's share increases from 10.6% to 12.5%.
By contrast, the Northern Cape shrinks from 30.5% of land area to 6.0% of agricultural output. Its difference of −24.6 percentage points makes it the largest contraction for Agriculture. What looks like a dominant province on the geographical map becomes a relatively small part of the agricultural cartogram.
Mining: a different South Africa emerges
Mining shifts the emphasis again.
North West and Limpopo each contribute approximately a quarter of the national mining total, with Mpumalanga also accounting for a substantial share. The Northern Cape contributes about 11.9%, much more than its population or finance share, and consequently retains more area in this cartogram than in most of the others. Gauteng, so dominant in population and finance, becomes relatively less prominent.
The result is a different economic geography, one centred much more strongly on the provinces associated with mineral extraction.
Finance: Gauteng dominates
Finance produces perhaps the clearest contrast with the original map.
Gauteng accounts for approximately 42.6% of national finance, real estate and business services output. Yet it occupies only 1.5% of South Africa's land. That makes the province visually dominant in the finance cartogram, while the Northern Cape, with approximately 1.3% of the industry's output, contracts almost completely.
This is the largest contrast between geographical footprint and economic contribution among the selected measures.
The Northern Cape is always the largest contraction
One result surprised me when I added the dynamic KPI cards.
Regardless of whether the selected measure was Population, Agriculture, Mining or Finance, the Northern Cape remained the province with the largest contraction.
At first I wondered whether the DAX was incorrect. But the figures explain it. The Northern Cape starts with such a large share of the country's land that none of its shares of the four selected measures comes close to matching it.
Even mining, where it contributes almost 12% of the national total, leaves it more than 18 percentage points below its geographical share.
The repeated result isn't a problem with the visual. It's one of the clearest findings in the report.
Design choices
A few decisions made the report easier to understand.
- Two maps, not one. Keeping the original geography visible avoids asking readers to remember the shape of South Africa while looking at the distorted version.
- Consistent province colours. Colours provide a link between the two representations even when the outlines change significantly.
- Full province labels. I considered abbreviations, but the full names are more useful for readers unfamiliar with South Africa's provinces.
- A table instead of another chart. The percentage-point difference explains the resizing without introducing another visual encoding.
- Two dynamic KPI cards. Highlighting the largest expansion and contraction gives readers an immediate starting point.
- No conventional basemap. A normal geographical basemap behind the cartogram would imply that its coordinates represent real places.
The objective was to make the changes in shape the main event, with everything else helping to explain them.
What I learned from Day 15
1. A map's area can tell the wrong story. A geographically large province is not necessarily the most populous or economically significant. Choosing area as a visual encoding is itself an analytical decision.
2. Cartograms need a reference. A distorted map on its own is interesting, but placing it beside the original geography makes the transformation understandable.
3. Check the geometry, not just the picture. The same total cartogram area and proportional provincial areas matter more than matching visual zoom settings.
4. A small calculation can explain a complicated shape. Subtracting land-area share from statistical share gives a simple measure of expansion or contraction, even when the polygon outlines become extremely distorted.
5. Repeated results can be findings. The Northern Cape appearing as the largest contraction for every measure initially seemed suspicious. Checking the data showed why it was correct.
Final thoughts
The Day 15 theme was Inside Out, and a cartogram feels like a natural way to explore it.
Nothing about South Africa's provincial boundaries has changed in reality. The population still lives where it lives, the farms and mines are still where they are, and businesses haven't moved because I selected another metric. What changes is how the map allocates its visual space. That can be a useful way to challenge assumptions, particularly when conventional maps give large geographical regions more prominence than their statistical contribution warrants.
For anyone working with sales territories, customer demand, economic activity or service provision, the lesson applies beyond South Africa. A region's physical size might be the least interesting thing about it.
The geography hasn't changed. What we're measuring has.
Sources and attribution
- Population: Statistics South Africa, Mid-year Population Estimates 2025 (P0302). Provincial estimates for 2025, national total approximately 63.1 million.
- Provincial economic output: Statistics South Africa, Provincial Gross Domestic Product, P0441.2, September 2026 release, published October 2026. Annual provincial industry figures for 2025, at current prices.
- Provincial boundaries: South Africa Subnational Administrative Boundaries (COD-AB), UN OCHA, distributed through the Humanitarian Data Exchange (HDX). Source GeoJSON:
zaf_admin1.geojson. - Cartogram generation: go-cart.io. Cartograms generated using go-cart.io from the original provincial boundaries and 2025 statistical values. Attribution guidance.
- Cartogram algorithm: Gastner, M. T., Seguy, V., and More, P. (2018). Fast flow-based algorithm for creating density-equalizing map projections. Proceedings of the National Academy of Sciences, 115(10), E2156–E2164. https://doi.org/10.1073/pnas.1712674115.
- Cartogram web application: Shi, T., Duncan, I. K., and Gastner, M. T. (2019). go-cart.io: A web application for generating contiguous cartograms. https://doi.org/10.5194/ica-abs-1-333-2019.
Methodology note: The report contains two map representations. The Geographic view uses the original administrative level 1 GeoJSON boundaries. Four contiguous cartograms were generated independently in go-cart.io using 2025 provincial population estimates and agriculture, mining and finance industry output. Each cartogram was exported as JSON, matched to provinces using ZA1–ZA9, converted to WKT and uniformly transformed into a display coordinate range for Icon Map Pro. The four statistical cartograms contain 36 province-metric records. No additional polygon simplification was applied. Provincial statistical shares are each value divided by the national total for that metric; land shares use each province's area divided by the sum of provincial land areas. Differences are expressed in percentage points, not changes over time. The cartogram geometries use artificial coordinates and do not represent geographical locations or distances. Population figures are estimates, and industry output is reported at current prices.