30 Day Map Challenge 2026 - Day 11 (Icon Map Pro) - Sound

What does South Africa sound like at home?

For Day 11 of the 30 Day Map Challenge, the theme is Sound. After Day 10's prompt-only pharmacy map, I wanted a sound you can map without a microphone. The most familiar sound of any place is the language people speak at home, and few countries have as many official languages as South Africa: twelve since South African Sign Language joined the list in 2023.

That gives the map a practical question for anyone who markets, sells or serves customers here: how many languages does your business need to speak to reach 80% of people in their home language? The answer changes from one language to seven depending on where you are, and the places that need the most are the places with the most customers.

Explore the interactive report

The map shades every area by its largest home language. The shading is solid where that one language alone reaches 80% of residents, and paler where it takes two, three or more. Pale areas are where the language mix gets complicated.

  • Show by: switch between Province, District and Municipality. The report opens on the 52 districts.
  • Hover: each area's card shows its five largest home languages as a small bar chart, and how many languages it takes to reach 80%.
  • Click an area, or lasso several, to update the cards and the national language chart for your selection.
  • Click a language in the chart to map where it is spoken: the map recolours to that one language, stronger where it is a bigger share of residents.
  • The table lists the areas that need the most languages. Click a row to see its language mix and find it on the map.

Interactive report: Day 11, The Sound of Home

What the map doesn't show: which language people prefer for business, or which language they understand. Many South Africans speak multiple languages, and home language is only the first of them. "80%" means the home languages of 80% of residents, not 80% of customers who will respond.

Starting with Census 2022 home language

The data is the Census 2022 question on home language, exported from Statistics South Africa's SuperWEB2 tool. I exported two tables: one for all 213 municipalities (8 metros and 205 local municipalities), and one with province, district and municipality rows that reconcile at every level. Each row has 17 languages plus "not applicable" and "unspecified", and the national total is 62,027,503 people.

A few things needed care:

  • Names don't match. Census names are written several ways ("Local Municipality of Madibeng", "Ethekwini" and "eThekwini"), and Emalahleni exists twice, once in the Eastern Cape and once in Mpumalanga. I matched on name and province to the Municipal Demarcation Board codes.
  • Shares, not counts. The census counts are weighted estimates, and the 2022 census had an undercount of about 31%, worst in the Western Cape. Shares are more robust than totals, so the map uses shares. They exclude "not applicable" and "unspecified", which range from about 2% to 15% by municipality.
  • Display names. I used the names the languages use for themselves: isiZulu, isiXhosa, Sepedi, siSwati, Tshivenda, Xitsonga, isiNdebele.

Stats SA's terms ask for the source to be acknowledged, for the analysis to be identified as the author's own, and for the data not to be sold. The report and this post do all three.

Boundaries without a GIS

For consistency with Day 8, I used the OpenStreetMap administrative boundaries I'd already extracted: metros and districts at admin level 6, local municipalities at admin level 8.

OSM still holds some municipalities that were merged in 2016, and there are no province outlines that match the census areas exactly. Both meant merging polygons, and I wanted to do it without a GIS library. OSM's own topology made that possible. Neighbouring areas share the same boundary ways, so when several areas are merged, every way that appears twice is an internal edge. Dropping those and joining the rest back into rings gives the merged outline.

Simplifying without slivers

The full-detail outlines were too heavy for a report. Simplifying each polygon on its own would leave thin gaps between neighbours, because each side of a shared border would be simplified differently. Instead I simplified each boundary way once, keeping its end points, so both neighbours use the identical simplified edge. Tolerances are about 400 m for municipalities, 600 m for districts and 1 km for provinces.

How many languages to reach 80%?

The key measure counts languages largest first and stops when they cover 80% of residents. In DAX it is a cumulative count. For each language, it checks whether the languages bigger than it still add up to less than 80%. If they do, that language is needed.

Because it is a measure, not a stored column, it works for any selection. Lasso the whole of Gauteng and the cards show that 7 languages are needed. Select the Eastern Cape and the answer is 1. Nationally it takes 7 languages to reach 80% and 8 to reach 90%.

One borderline case shows how sharp an 80% line is. isiZulu is the home language of 79.96% of residents in KwaZulu-Natal, so the province counts as needing two languages. It shows in the paler shading. I've left the rule as it is: a threshold has to fall somewhere, and the card shows the real share.

Bringing it together in Icon Map Pro

One map, three levels

All three levels sit in one table with a level column, and the Show by tiles filter it. Because the census totals already reconcile at every level, there is no aggregation in DAX: each level is its own set of polygons with its own numbers.

Colour for the language, transparency for the mix

The fill colour is the area's largest language. The fill transparency comes from a measure on the number of languages needed: solid for one, 35% transparent for two, 55% for three, 65% for four and 75% for five or more. Clicking a language in the chart changes both measures, so the map switches to that language's colour with transparency based on its share.

The colours follow the language, not the area. isiZulu is green, isiXhosa orange, Afrikaans violet, Setswana yellow, Sesotho brown, Sepedi light green. The chart beside the map uses the same colours, so it doubles as the legend.

SVG charts on the hover cards

Icon Map Pro's card labels can show an image, and the image can be an SVG built in DAX. The Label chart measure takes the area's five largest languages, draws one bar per language in its colour with the share beside it, and returns the result as a data URI.

Two fixes made the cards work:

  • Sharp text. The first cards looked fuzzy, because the image was drawn at exactly the size it was shown and then scaled for the screen. Drawing the SVG at three times the size, with a viewBox that keeps the layout, lets the card scale it down instead of up, and the text stays crisp.
  • Ties. In some areas two small languages both round to 0%. They got the same rank and were drawn on the same line, and the "top five" became six. A tiny tie-breaker based on each language's position in the list gives every language its own row.

The Power BI tooltip is switched off, so the Icon Map card is the only pop-up, and the card title stays black for legibility. Yellow Setswana text on white is hard to read, and the first bar already shows the language colour.

What the map reveals

Most of the country needs one language. In 119 of 213 municipalities, a single home language reaches 80% of residents. isiZulu is the largest language in 55 municipalities, Afrikaans in 46, isiXhosa in 35 and Setswana in 29. Maphumulo, Msinga, Nongoma and Nkandla in KwaZulu-Natal are about 99.5% isiZulu.

The customers are where the mix is. The 119 single-language municipalities are home to about a third of South Africans (21 million people). The 15 municipalities that need four or more languages hold a quarter (about 16 million). Johannesburg and Tshwane need seven, Ekurhuleni and Mogale City six, and Rand West, Madibeng and Emalahleni in Mpumalanga five. Tshwane is the most mixed: Sepedi is largest at 24%, followed by Setswana at 18%. In the other cities, Cape Town needs three languages (isiXhosa 37%, English 31%, Afrikaans 26%) and eThekwini two.

English is everywhere, and the largest nowhere. English is the home language of 5.2 million people, fifth nationally, but it isn't the largest home language in any municipality. It is second in 57 municipalities, including Johannesburg, Cape Town and eThekwini. A national campaign in English reaches almost everyone who reads it, but it doesn't speak to almost anyone in their home language.

These are home languages from a census with a large undercount. They describe what people speak at home, not what they prefer to be addressed in, and they don't measure how many languages each person understands.

What I learned from Day 11

1. Count to a threshold, not just the largest. "Largest language" makes every area look settled. "Languages needed for 80%" shows where the work is.

2. Topology can replace a GIS. When neighbours share boundary ways, merging polygons is a counting exercise, and simplifying the shared ways once keeps borders clean.

3. Draw SVG images bigger than you show them. A viewBox at three times the size costs nothing and keeps text sharp when the visual scales the image.

4. Plan for ties in any top N. Rounded shares tie more often than you'd expect, and a top 5 with ties isn't five rows.

5. One map, many questions. Recolouring the same polygons by the selected language turns a summary map into a map for each language.

Final thoughts

A map of the largest category is the easy version of any mix: the biggest party, the main crop, the most common product. Counting how many categories it takes to reach a threshold tells you how much effort an area needs, and that is usually the business question. The same measure works for customer segments, product ranges or service languages in your own data. Shade by the largest, fade by the effort, and let a click show where any single category is strongest.

Sources and attribution

  • Home language: Statistics South Africa, Census 2022, home language by province, district and municipality, extracted from SuperWEB2, November 2026. Source: Stats SA; the analysis and processing are the author's own. Used under Stats SA's terms of use; the data is not sold or included in a paid product.
  • Boundaries: © OpenStreetMap contributors, available under the Open Database Licence (ODbL). Geofabrik South Africa extract, 30 October 2026. Municipal codes from the Municipal Demarcation Board, as tagged in OpenStreetMap.

Methodology note: Shares are home-language speakers divided by all residents who gave a home language, so "not applicable" and "unspecified" are excluded. "Languages needed for 80%" counts languages from the largest down until they cover 80% of residents in the area or selection. Boundaries are OpenStreetMap admin level 6 (metros and districts) and level 8 (local municipalities), matched to census areas through Demarcation Board codes. Pre-2016 municipalities are merged into their successors, and provinces are built by merging municipalities. Shared boundary ways are removed when merging and simplified once each (about 400 m, 600 m and 1 km tolerance for municipalities, districts and provinces). The Prince Edward Islands are excluded, and Collins Chabane (LIM345) has no OpenStreetMap outline. Census 2022 counts are weighted estimates, with an undercount of about 31%.