Would a few thousand unmatched/tied NYC addresses be a source data issue?

Question

Background: Am geocoding thousands of addresses, a process I am quite a novice too, using the NY State address composite locator. Using ArcMap 10.3.

14% remain tied and 12% unmatched. I'm inclined to think this is a source data issue but because of my inexperience I can't be sure.

A few notes:

1. The unmatched tied addresses are almost exclusively hyphenated addresses. But many of the addresses that were successfully matched are hyphenated as well.

2. It seems a handful of streets are the worst offenders for unmatched/tied addresses.

3. None of the matched addresses have a score below 100.

Answer 1

This is not uncommon. New York is the most difficult region to obtain matched addresses to a location. Even best geocoding engines struggle to provide match better than .7 There are few reasons, one is that geocoding engines still are not applying previous matches as IA, but most important that New York has very difficult address structure: state and city are matching and street names are just ugly. Then you need to look if secondary address unit is present. That will bring down accuracy in New York to less than .7.

To be successful, please wire a two or three more backup systems to run over again the failed locations until you get them to 100%

Disclaimer: I work for batch geolocation service company at CSV2GeoData