Most of my mapping is done via “armchair mapping”, using aerial imagery to add buildings and features around. I’m thankful that where I live, I have high quality, and relatively recent aerial imagery. Because most of what I map tends to be residential (since in order for the map to be useful to people, they want to know that where they live is actually present), I find that I add a significant number of houses, garages, and sidewalks/footpaths. When available, there are open data sources for addresses, so I’ve found that adding data in stages seems to help.
The first stage being correcting roads that haven’t been touched since the Tiger imports, fixing and adding land use such as residential or retail or farm areas, adding paths and other routing features.
The second stage I add buildings such as houses, retail, and whatnot, fix imported church and post office/official building locations based off internet search addresses and educated guesses (e.g. does the building look like a church or have white vehicles nearby? Then it’s probably the nearby labeled church or post office, etc. Also does searching for {town name} and eg “baptist church” etc yield usable results like a website? Sometimes not, but most often yes).

In addition to buildings, I like to add power poles, fire hydrants, and other service things like benches, street lamps, etc if they’re present and I can see them on the aerial imagery (or can assume they’re there from wire locations and other context clues).

For stage 3, I like to add addresses, driveways and the power lines between the poles. Addresses I can only add if I know of an open data source (where I’ve been mapping, addresses are freely available. In the US, generally the county website will have a GIS app that may or may not have an open license for the data it contains. If someone smarter than me is able to convince more counties to offer that data for free, that would be an incredible boon to address mapping in the states, as well as being able to import them. I’m unfortunately not skilled enough with JOSM or other more advanced tools to perform that).

I generally like to add addresses and driveways at the same time since using the iD editor there is no easy way to tell if a building has an address or not in its tags (beyond mapping the addresses as separate points to the building, but I believe that is not a good convention to follow), having the visual confirmation of the driveway indicating a present address helps, it also aids in routing since it can connect sidewalks and the main street, providing a more accurate reflection of reality. During this stage, I also look for shadows at crossroads for stop signs. Generally ESRI maps are taken in the morning or evening, so stop signs and other street signs cast shadows that can be seen, and if the street has a four way stop, there is a little sign underneath the large octogon of the stop sign that indicates “4 way”, so that determines the type of stop to use at the intersection. Additional details can come much later such as building colors and other data that can be inputted from StreetComplete or other map apps.
Ideally, the map should end up reflecting reality as much as possible. Here’s an example of an area that I micromapped and is fairly close to reality when comparing between Google Maps’ 3D view and F4 Map’s 3D view.

This is close to the state that I would like to see the whole world in, but unfortunately there’s significant areas missing data, so for the majority of places, we’re still stuck in stage 1.