PERSONAL INTERFACE — Portfolio
Built to be checked
Things built here. The first is a cartogram, shipped with its values missing — an empty layer is better than a number nobody can check.
A tile cartogram of Korea
A tile cartogram draws every unit at the same size. On a real map Seoul is a dot and Gyeongbuk is enormous, so colouring the true shapes makes the reader compare areas when the question was about the value. Equal tiles remove that.
Registered population
As of 2026-07-31. It counts registrations rather than people — it includes residents whose whereabouts are unregistered and excludes foreign nationals. The published file is 3,619 rows of neighbourhoods with no province totals in it, so these numbers are those rows summed by province.
16 divisions · bar length is population.
- Gyeonggi · 13,768,157
- Seoul · 9,284,263
- Busan · 3,230,982
- Gyeongnam · 3,193,742
- Jeonnam–Gwangju · 3,157,777
- Incheon · 3,063,730
- Gyeongbuk · 2,494,470
- Daegu · 2,347,389
- Chungnam · 2,138,955
- Jeonbuk · 1,717,841
- Chungbuk · 1,601,156
- Gangwon · 1,507,224
- Daejeon · 1,442,190
- Ulsan · 1,086,756
- Jeju · 662,680
- Sejong · 390,972
Brokerages per 10,000 residents
Drawing the raw office count would redraw the population map — 82,445 nationally, and the largest counts sit where the most people do. Dividing changes the order. Sejong at 21.5 and Jeju at 21.0 come out above Seoul at 20.4, and Gyeonggi, which has more offices than anywhere else at 22,708, lands mid-table at 16.5. Gangwon is last at 11.4.
16 divisions · bar length is offices per 10,000 residents.
- Sejong · 21.5
- Jeju · 21.0
- Seoul · 20.4
- Gyeonggi · 16.5
- Busan · 16.2
- Daejeon · 16.0
- Daegu · 15.4
- Incheon · 14.7
- Gyeongnam · 14.5
- Chungnam · 14.4
- Ulsan · 14.3
- Jeonbuk · 14.1
- Jeonnam–Gwangju · 14.1
- Chungbuk · 13.1
- Gyeongbuk · 11.8
- Gangwon · 11.4
The renewal right and the 5% cap →
Direct deals, and the better explanation →
Apartment price change · by district →
Apartment management app revenue →
Revenue dashboard · filtered →
Why it stops here
- ~250 districts
- The tiles would fit the budget. The positions would not: they are not checkable at a glance, and I do not have real centroids under a license that permits redistribution. Placing 250 tiles from memory would be inventing data.
- ~3,500 neighbourhoods
- Measured: this route may spend 23.4 KB of gzipped HTML, and 3,500 tiles do not fit. A 3,500-tile cartogram is also not one anybody reads.
- The value layer
- Still empty. It has to be 공공누리 type 1 — a visualisation is a derivative work, and types 3 and 4 forbid modification — its labels have to work in both locales, and it has to answer a question rather than exist. No dataset has cleared all three yet.
The plan this page executes against is docs/DATA-VISUALIZATION.md. One item on it is finished — this grid. The rest has no evidence yet.