Every city centre has cafés, pharmacies and banks, but not in the same proportions. Counting every listed business within 5 km of the centre of 49 cities, and comparing each city's mix with the average of all of them, gives each an economic fingerprint.
FindingRestaurants per 1,000 listed businesses range from about 300 in central Seoul to about 35 in central Lagos.
How to citeAl, Stefan. 2026. “What Cities Sell.” Formlume. https://research.stefanal.com/city-business/
DataOverture Maps Foundation places (CDLA-Permissive 2.0). Please also cite the original data when you use these figures.
UseFigures on white may be reused for non-commercial purposes with credit (CC BY-NC 4.0). The page design is © Stefan Al, all rights reserved; for commercial use, please get in touch.
Business types this city has most of relative to the 49-city mix (types found in at least half the cities, with 60+ listings here), as a multiple of the average share.
Common types (among the 60 most frequent overall) this city has least of.
Listings of this type per 1,000 businesses within 5 km of each centre.
Source. Overture Maps Foundation places (CDLA-Permissive 2.0), which merges business listings from Meta, Microsoft and others, read with the Overture command-line tool. Listings with a confidence score under 0.5 are left out; each is counted under its primary category.
Cities. Addis Ababa is left out: fewer than 2,000 places are listed there, too few for a stable mix. Places include shops and services but also churches, parks and public offices, so "business" is used loosely.
Measure. For each city, the share of its listings in each category, divided by that category's average share across the 49 cities (a location quotient): 2.0 means twice as common as usual. The same fifty cities and centres as the density cones and building-mass pages.
Reading it. Listings reflect who puts businesses online, and how: platforms are stronger in some countries than others, informal trade is largely missing, and categories are assigned by machine. Read the distinctive extremes, not small differences.
Tools. Python (the Overture Maps CLI, GeoPandas, pandas) and D3.js.