Slightly drifting off the Overture topic here (and it may be worth breaking this out into a separate thread), but the TIGER data in OSM is being actively fixed and it is possible, with some work, to get good results from it. Yes, if you treat it naively the results won’t be great, but that’s true of all OSM data. cycle.travel applies a few heuristics to it, and the result (IMobviouslybiasedO) is significantly better cycle routing in the rural US than Google manages. If I can do that with one of me working part-time from 3500 miles away, I’m sure Facebook and Microsoft can manage.
The TIGER-sourced data is obviously still a long way from perfect, and personally I think a semi-automated approach to clear up the mistakes of an automated import like TIGER would be 100% fine. I’m interested to see how the imagery-derived surface tagging approach in New algorithm for inferring road surface tags pans out. I also wonder if some of the data being released by Overture may end up being a useful input here. There’s also some useful state-level data that could be brought into service.
But POIs are arguably our biggest unsolved challenge in the rural US right now IMO, not TIGER.