Country charts stopped being useful years ago

Five breakdowns that country charts cannot give you

Regions come with the ISO subdivision code and the GeoNames identifier, so “Bavaria” and “Bayern” are the same row rather than two. Cities carry coordinates and an accuracy radius, which is what makes a map honest: a point with a twenty-kilometre radius is a range of addresses, not a person, and the interface says so.

Networks list the autonomous system number, the provider name, the proxy type of that range and — crucially — human clicks separately from total clicks. That single column answers the question every media buyer eventually asks: which provider is sending me traffic that never converts.

Traffic quality splits visits by visitor type, by proxy type and by quality score band. And activity by local hour is computed in the visitor’s own time zone at the moment of the click, because a storage engine can only convert to one fixed time zone and every row here has a different one.

• ✓Regions with subdivision codes and GeoNames identifiers
• ✓Cities with coordinates and an accuracy radius
• ✓Networks with autonomous system, provider and proxy type
• ✓Visitor type, proxy type and quality score bands
• ✓Activity by the visitor’s local hour, not by yours

Place names in the reader’s language, not the writer’s

A click event stores an identifier, never a localised name. The name is resolved when the report is read, from the dictionary of the language the reader is using — which is why the same click reads as “Berlin” for one colleague and “Берлин” for another, from one stored row.

Names are available in eight languages out of the ten the dashboard speaks; the remaining two fall back to English rather than to an empty cell. The dictionaries are megabytes and live server-side: nothing of the sort is shipped to a browser, and the locale is taken from the reader’s own setting rather than from a request parameter — a parameter would let one person poison the shared cache of a workspace with someone else’s language.

What the data is, and what it is not

Geo answers come from licensed commercial databases packed into a format built for the edge, with a documented fallback: if the pack is unavailable the resolver degrades to the network-level data the CDN already provides and marks the result as approximate. Every row records which source answered it, so a chart never mixes precise and approximate data silently.

The visitor’s address is never stored. It exists inside a single function long enough to produce a daily fingerprint — a salted hash rotated every night — and to look up the address range; it does not reach the event. Postal codes are resolved but deliberately not recorded: in dense housing a postal code is close to an address.

Place data is credited where credit is due: city and region names come from the GeoNames database, used under CC BY 4.0, and the attribution is published on the site rather than buried in a contract.

Detail is a setting, not a fixed price

Every workspace decides how much of this to keep. Full detail stores coordinates and time zone; “city without coordinates” keeps the market and drops the point on the map; “country only” keeps the continent and country and nothing else — and at that level the geo database is not even queried for the event.

This is a setting about writing, not about reading. What a privacy mode turns off is never written down, so no report, no export and no future query can bring it back. Retention works the same way: each event carries its own retention, and the storage engine deletes the row when it expires.