* someone knows how to use a map to a very high degree of precision (not true often)
* GPS is accurate enough, and this is a double whammy - on initial assignment of the GPS point (i.e. it's the GPS of their door/gate, not of their living room couch, and it was collected with enough precision) and on usage (i.e. the GPS antenna on a $30 smartphone)
* plenty of other issues
There are a couple startups trying to solve this beyond what3words mentioned in other comments - they'd suffer from many of the same issues I mention above. Some encourage people to put a placard outside their door/gate with the code to make last-mile arrival easier.
But a more successful approach from research I've done is what OkHi (http://okhi.com) is building - use a photo attached to GPS. From n>1000 tests, we found that the photo drastically decreases time over the alternative, even reducing average time to find the location by over 50% in the last 500m for some addresses.
(Full disclosure - I was worked on the addressing problem for the two years in Nairobi at OkHi, so I obviously believe in their solution or wouldn't have dedicated that much time there :)