I've been struggling with creating new style locators, and publishing them blah,blah,blah for a while, but I just came upon something that is really cool, and I consider it a win.
To test new locator functionality, I got a couple thousand addresses from a permitting data base here in the county. As usual, the addresses were pretty ugly, but after a quick visual scan, I decided on a quick python clean up routine and went about geocoding them.
After clean up, one the address looks like this:
IN-STORE (WHOLESALE) HARMONS 4874 WEST 6200 SOUTH KEARNS UT
which in my view is about as bad a single field address to match as you can get. But....
My geocoding process returned a 61.25 score against a point address that has the address of 4874 W 6200 S. That's pretty cool in my book. This turns the tables quite a bit for me as I generally don't consider anything under about 85% as reliable.
That should just about do it....
