Where do South Africans live, and where might access to supermarkets be limited?
For Day 1 of the 2026 30 Day Map Challenge, the theme is Points. I wanted to go beyond plotting locations on a map and explore how two different point datasets could help answer a practical retail question.
South Africa provides an interesting setting. Its population is unevenly distributed, with major concentrations around Johannesburg, Pretoria, Durban and Cape Town, alongside vast areas with relatively few residents.
But how does that population distribution compare with supermarket locations?
Using the Icon Map Slicer visual, I created an interactive Power BI report combining estimated population figures with supermarket locations from OpenStreetMap. Users can explore population concentrations, zoom into individual towns and cities, and draw a circle to investigate the population and competing stores within an area.
Explore the interactive report
Zoom into a city to reveal supermarket locations ( from zoom level 8 ), then draw a circle to update the population and supermarket figures. The legend explains the proportional population circles and the store markers.
Here's the report to play with:
Interactive report: Day 1, A Nation of Places.
Two datasets, one geographical question
The map uses two openly available datasets.
Population: Global Human Settlement Layer
The Global Human Settlement Layer (GHSL) is produced by the European Commission's Joint Research Centre. Its GHS-POP dataset provides gridded estimates of residential population.
For this map, I used the 2025 population projection at 30-arcsecond resolution, downloaded as GeoTIFF tiles covering South Africa. Each raster cell contains an estimated population figure. Rather than displaying these values as a conventional raster, I wanted to represent the data using points, in keeping with the day's theme.
Supermarkets: OpenStreetMap
Supermarket locations came from OpenStreetMap, extracted using Overpass Turbo. The query selected features tagged shop=supermarket, including nodes, ways and relations, with their available names, brands and coordinates.
The extracted dataset contained 2,804 supermarket records. This is a count of mapped records, not a verified complete inventory of South African supermarkets. Coverage depends on how current and complete the contributions are.
Turning population rasters into points
Preparing the population data presented the first challenge. The GHSL GeoTIFF tiles were processed using Python to extract the population value and centre coordinates of each populated raster cell.
This produced 536,395 population cells, far too many to display in Icon Map alongside thousands of supermarkets.
My first approach reduced the dataset to approximately 25,000 population-weighted points. Although this substantially reduced the number of features, the map was still crowded, with overlapping circles obscuring the underlying geography. The solution was to aggregate further, producing approximately 5,000 weighted points.
Rather than randomly sampling locations, nearby population cells were grouped together. Each new point was positioned at the population-weighted centre of its contributing cells, while the population values were summed. This preserved the input population total while reducing the number of map symbols.
The result was a more legible national map. Major population concentrations remained prominent, while sparsely populated areas became much less visually dominant.
There is an important trade-off: these points represent aggregated population, not individual settlements or precise residential locations. Their coordinates are representative centres, so small catchment selections produce approximate population totals.
Using proportional circles to communicate population
The population layer uses blue proportional circles, with each point representing the estimated population of its contributing GHSL cells. I experimented with circle size ranges, transparency and square-root-based scaling. The objective was to make larger population concentrations visible without allowing the biggest circles to dominate the map.
Adding supermarkets without overwhelming the map
The second point layer contains supermarket locations. I created simple SVG markers using DAX, assigning different colours to supermarket brands without using their logos.
However, displaying all 2,804 supermarket markers simultaneously overwhelmed the national view. I therefore used scale-dependent visibility: population circles remain visible at national scale, while supermarket markers appear at zoom level 8 and above. At a national scale, the map communicates population distribution. Zooming into a town or city reveals the individual supermarket locations, while keeping the focus on the Points theme rather than introducing clustering or hexagons.
Exploring retail catchments with circle selection
Icon Map Slicer allows users to select geographical features by drawing a circle on the map. At zoom level 8 or above, a circle around a potential retail location updates three KPIs:
- Estimated Population: Population represented by the selected weighted points.
- Supermarkets: Number of mapped supermarkets included in the selection.
- People per store: Estimated population divided by the number of selected supermarkets.
A horizontal bar chart shows supermarket counts by brand for the selected area.
For the final report, I retained the zoom threshold because it improves readability. The report instructions explain that users should zoom to level 8 or higher before analysing competing supermarkets.
This is a reminder that cartographic design can affect not only what readers see but also how they interact with data.
What does the map reveal?
The finished map makes South Africa's uneven population distribution easy to explore. Large concentrations appear around its major urban centres, while much of the western interior is comparatively sparsely populated. Among records in the extracted OpenStreetMap dataset, Pick n Pay, SPAR and Shoprite are among the most frequently represented supermarket brands.
The people-per-store ratio offers a starting point for spotting places that might deserve further retail investigation. It is not a definitive measure of unmet demand. A proper retail assessment would also need to consider travel time, road access, store capacity, purchasing power, customer behaviour and possible gaps in OpenStreetMap coverage.
What I learned
More points do not necessarily make a better map. Reducing more than half a million population cells to roughly 5,000 weighted points made the national distribution considerably easier to interpret.
Symbol sizing matters. Small changes to size, scaling and transparency substantially affect how the geographical patterns read.
Scale-dependent visibility improves clarity. Showing supermarkets when zoomed into a city keeps the national map readable.
Open data needs context. GHSL values are modelled estimates, the population points are aggregated, and OSM supermarket coverage may be incomplete.
Final thoughts
Day 1 demonstrates that a point map can do more than show where things are. By combining proportional circles, SVG location markers, scale-dependent visibility and spatial selection, Icon Map Slicer helps readers explore relationships between population and retail locations directly inside Power BI.
The challenge was not getting more points onto the map. It was finding the right balance between detail, readability and analytical usefulness.
Data sources and acknowledgements
Population: Schiavina, M., Freire, S., Carioli, A. and MacManus, K. (2023). GHS-POP R2023A: GHS population grid multitemporal (1975–2030). European Commission, Joint Research Centre. Dataset DOI. The 2025 projection is available under CC BY 4.0. Gridded data was converted into population-weighted points for this report.
Methodology: Pesaresi et al. (2024). Advances on the Global Human Settlement Layer by joint assessment of Earth Observation and population survey data. DOI.
Supermarkets: © OpenStreetMap contributors, available under the Open Database Licence (ODbL). Locations extracted using Overpass Turbo in October 2026.