> town

That's where the problem lies. I spend too much of my life screaming into the void about UK addresses and what is and isn't a town :)

It's just about possible to create dataset of towns, cities, villages from mixing and matching and cutting up other geographies from the ONS (e.g. their Parished and Unparished data (though Unparished is its own nightmare), then the ONS Built-Up Areas dataset but... that one cuts out all the non-built parts so Milton Keynes is bunch of big blobs around the residential areas).

But if you frankenstein it, you lose all of the formal connections that define other links, and if you try to cheat via geospatial join, you will link police forces to areas they don't actually serve.

For Swindon, you're kinda stuck glueing together (going from memory here - please correct me if I'm off) the Parishes of Central Swindon, South Swindon and West Swindon. But these will overlap with some villages so it's not a true count.

And that's a natural moment to lead on to the next critical point: how're we measuring population? (I always get that Father Ted speech scene in my head thinking about this, start speaking - then twelve hours later "And now, we move on to liars")