Matt,
Geocoding is never an exact science nor can it possibly be as accurate
as
the actual obtained coordinates for a client. Geocoding first relies on the
accuracy of the Census Tiger line mapping databases. All mapping companies
use this as a base and then do various levels of data checking and
refinement to improve the results. Important facts to note about geocoding:
- The result will only resolve to the road frontage not the structure.
- 911 Addressing is supposed to be standardized so that every 50 feet or so
of road frontage has an actual numeric address. If a locality has deviated
from the standard you will get inaccurate results from some mapping sources.
- Tiger map data knows the range of address numbers between intersections
and will estimate the point along the highway based on the 50 foot rule.
- Some geocoding software will tell you the accuracy for which it resolved
the address record.
- For different parts of the country different companies have different
levels of accuracy based on their efforts to improve the data.
- There are only three or four major companies that do this type of work and
therefore most mapping companies will contract for that data. This is why
you can see the same errors from different sources. They all used the same
data source. GDT is one of the big sources.
When I geocode addresses using GIS software, I can get results back that
tell me at which level of accuracy I was able to achieve. It can fail to
create a point on whatever level I wish. The three basic levels are,
building match, street match which means it placed it on where the address
should be along the road, and then zip code match. For the zip code match it
will place the location at the default centroid point defined by that zip
code polygon. It will not always be the post office. the quality of how the
address data is formatted can make a huge difference as well. How people
abbreviate some things will cause wildly different results.
On the topic of zip codes, it is important to note that the post office does
not define polygons and zip code areas. That was something the Census Bureau
created. Zip codes are a linear routing function. If you would like a full
explanation with illustrations go to this link
http://www.manifold.net/doc/manifold.htm which is the software user manual.
Click on the index button and go to the "z" section and look at the topic
"zip codes are not areas"
The majority of the Geocoding problems are a result of the "garbage in
garbage out" syndrome of any database system.
Thank You,
Brian Webster
-Original Message-
From: wireless-boun...@wispa.org [mailto:wireless-boun...@wispa.org]on
Behalf Of Matt Larsen - Lists
Sent: Monday, April 06, 2009 10:51 AM
To: WISPA General List; w...@part-15.org
Subject: [WISPA] Bad Geocoding Data
This article in the LA Times:
http://www.latimes.com/news/local/la-me-geocoding-errors5-2009apr05,0,596628
5.story
This documents the reasoning for why I have not completed my Form 477
data yet. Nearly 40% of my customer base will have to be re-coded for
the Form 47 because the geocoding databases are incorrect. My lead
tech has exported the geocode data out of Freeside and into Google
Earth, sorted by AP. When we look at the data, a very high percentage
of our customers have GPS coordinates of Post Office of their
town/village. We still have a lot of county road and rural route
addresses in this area, and they don't geocode correctly.
Data with 40% noise borders on useless. I applaud the spirit behind
the 477, but asking us to provide this granular data without the right
tools to assemble the data and verify it makes it a nearly unanswerable
proposition.
Matt Larsen
vistabeam.com
WISPA Wants You! Join today!
http://signup.wispa.org/
WISPA Wireless List: wireless@wispa.org
Subscribe/Unsubscribe:
http://lists.wispa.org/mailman/listinfo/wireless
Archives: http://lists.wispa.org/pipermail/wireless/
WISPA Wants You! Join today!
http://signup.wispa.org/
WISPA Wireless List: wireless@wispa.org
Subscribe/Unsubscribe:
http://lists.wispa.org/mailman/listinfo/wireless
Archives: http://lists.wispa.org/pipermail/wireless/
