How we build a city dataset
A city page on InfoGap is not a blog post about a city. It is a dataset with a definition, a count and a date. This page is the method behind all 386 of them, so you can decide for yourself whether the numbers are worth citing.
The numbers
| Value | What it counts | Source / definition |
|---|---|---|
| 13,204 | place records | Counted from the 386 city dataset pages served at infogap.store/china/ on 2026-09-25. One record = one place. |
| 386 | city and region datasets | One page per city or region under /china/. Datasets rebuilt 2026-09-24; the hub counts them from disk, not from a list somebody typed. |
| 384 | pages that print their own record count | 2 of the 386 pages do not carry a count in their description. They are still served; they are simply excluded from the 13,204 figure rather than guessed at. |
The rule
- Collect. Place records come out of our own place database. Each row already carries a category, a coordinate, a rating and, where the source provided it, an address.
- Normalise. One row becomes one record. Rows without a usable coordinate are dropped rather than placed at a city centre — a wrong dot is worse than a missing dot.
- Translate, and keep the original. Every record carries the Chinese name next to the English one. We never replace the Chinese name: if our English is bad, you can see it and say so.
- Count. The number printed for a city is the number of records actually written into that page. If a record is dropped, the number falls by one. The site-wide total on the data hub is the sum of those page counts, recomputed whenever the hub is rebuilt.
- Publish with a date. Each page states when its dataset was last rebuilt. A number without a date is a claim; a number with a date is something you can check again next month.
The same method runs for every city, which is the only reason the pages can be compared with each other. A dataset that used a different rule for one city would make the other 385 less useful, not more.
Where it breaks
This method has three known failure modes. One: coverage is not uniform — a large city with a well-mapped database produces a long list, a small region produces a short one, and a short list is not evidence that the place is boring. Two: ratings travel with the source record and are not our editorial judgement. Three: translation errors exist, which is exactly why the Chinese name is kept in the row: treat the Chinese name as the record and the English name as a convenience. Where we find an error we fix the record and the count changes with it.
Data & methods
- Rule authored by. InfoGap.
- Definition. One record = one place that exists in our own place database: a Chinese name, an English name, a rating and the district it sits in.
- Last updated. 2026-09-25
- Licence. CC BY-NC 4.0 for the text — attribution: InfoGap (infogap.store), link to this page.
Where to go next
- Data pages: https://infogap.store/china/
- Hub: https://infogap.store/data/
- Machine-readable: registry.json