I’ll give a contrived example, say you’re a property startup, and you want to produce a walkability rating for each house you sell. You might want to do a calculation on OSM for restaurants within 5 minutes walk, maybe a chef rating for whether there’s a greengrocer, fishmonger, delicatessen and butcher within walking distance, or you want to take into account a walk back from the train station and the opening times to see whether the shops are still open after someone comes back from their commute. This is the kind of thing which would be interesting for big cities with a lot of commuting by public transport.
My understanding is if you did this with OSM data you’d either have to make the algorithm that did the calculation public, or the output dataset (locations as keys and the final ratings for each as values).
OSM has exceptions for horizontal slicing (layer separate data on top of an OSM basemap), for vertical slicing (union OSM data with separate data) and for trivial manipulations, but it seems it doesn’t if you do this kind of non-trivial processing of OSM data.
So that means that it’s almost impossible for a company to find an interesting way of processing the data which can give them an advantage over a direct competitor, and there’s no real incentive to do so.
This is my reading of the terms from a few months back, hopefully I’m wrong about it. But I suspect this is why it’s rare to see the data being used like this.