When driving to a destination, the street address does not always correspond to the street where you park. In the case of someone in a condo (like me), the garage entrance is at the side of the building, not at the front. The roads you take to the address is not the same as where you would go if you want the parking garage (some one way streets would require you go around the block).

While I could mark home as the side street, there are similar problems for retail stores, where because of divided roads, parking lots, multiple entrances, etc. the same problem pops up in other scenarios.

With all the data you have going to the address, you can create a "heat map" of hot spots where people actually go and park. From this user data that you gather, you can map out where people really go and park when driving to an address. You can take to top 50-75% common locations people go and based on the specific user's behavior, use the directions to where they most commonly park. You could default first time drives to the address, or if overwhelming a different location, use that location as the "address". You could also filter the location by mode of transport (if you add public transport features).

If you create a heat map, as the user approaches the address, you could display a small or split screen 2D heat map of where people typically park in a 2 block radius. This would be nice convenience feature for people going to new addresses to show where they might find good parking and if there is a pattern.