A good example aside from Walmart can be churches. I am not sure who exactly, but someone added a ton of small churches to OSM years ago.
For supermarkets and store layout mentioned, I feel the problem being alluded to is when a business is in a larger strip mall/shopping mall or other complex. You could place the pin on the huge building, and it’s correct that it is there somewhere. But you want the building plan to understand at least which quadrant or something in the building it is. Of course, some users successfully map even this, but it’s good to place the pin where it should be.
I think some people have different ways of seeing the “place location” as a single lon,lat (a place is almost every a singe lon,lat, often it is some size of polygon so that pin is never correct anyway). Some people seem to think that it should be at the doorway, or just inside it. That can make sense, but some places have multiple doors… eventually, we arrive at mapping entrances. I think entraces are often mapped without any association. The place location of a restaurant certain can be ambiguous, as long as it’s inside the right building and adjacent to relevant streets or footpaths or parking, I think. Then an entrance (main) should be added that is either a relation, or has some tag similar to “operator=” indicating which node you enter to. This is in fact how a professional company making POIs (like Google) would do it.
Overture I believe would also be built in such a way that an entrance dataset would be supplementary. OSM can do much better, even using a piece of Overture data, by correcting the lon,lat if needed, keeping useful attributes like address and phone number, then adding an entrance (and adding the building with its own address too if needed). Overture also delivers the unique key GERS which could be interesting for OSM to track for example if a GERS key has new attributes, or is deleted from Overture, to flag it in OSM (vice versa not workable I guess). I see a lot of amenities in OSM that Every Door highlights as being more than 7 years old, so is very likely junk/nonexistent, and using signals from external data can help intensify the risk there (if it does exist in Overture, then either both are wrong, or both right, vs disgreement where 50/50 chance one is wrong, all helpful). POIs are special of couse, as we don’t often worry as much if a building or road was last edited 7 years ago.
In another look at the Walmart example with multiple departments, Google Maps started this new “directory” feature where there could for example be a single large mall, then a directory of things in it–the shops–and if some larger shops have something again further within it, like a pharmacy or cafe, that can also be specified. It’s not easy to make that, but I think probably relies on a more knowledge graph-based setup similar to Wikidata than to OpenStreetMap.
