Field Checking the Google-Base
The other day I had an appointment in Newport Beach, CA (yes, it is a tough life, but somebody had to visit the suntanned set). The address was 510 Superior Ave and I was amused that the proprietor of the business chose to give its position as “on Superior between Placentia and 17th Street, a distance that spans about a mile, with 5 streets intersecting Superior between the two endpoints. Why so vague?
To resolve this ambiguity, I looked at Google Maps and found to my surprise that Google had the location quite a distance south of the general area indicated by the business owner. I tried the maps at Yahoo, Bing and several others and all suggested that the 510 Superior address was north of where Google placed it and all located it in approximately the same place. What to do? Sounds like field research to me. As it turns out, Bing, Yahoo and MapQuest were right and Google was wrong. See figure below.
Sometimes wrong is even… wronger! Yep, the route to this location is a howler and one that makes it difficult to get back to the correct location, especially at peak traffic times.
The very next day friend of mine from Chicago sent a link to an article in the Chicago Tribune Business section titled “Google Maps inexplicably mixes Chicago’s past, present.” The article was another howler and evidence that Google Maps is now producing some of the best humor in the US (although, perhaps, not as interesting as Tiger). Google had positioned a college on the map at the location it occupied over 50 years ago and portrayed another college with the name it had 20 years ago, when it was acquired and its name changed. In addition, several neighborhood names and street names across the map were incorrectly labeled.
Somewhat humorously, the Google spokesperson contacted about the problem is quoted as saying that the information for Google Maps comes from a “wide range of public and authoritative data sources.” Good to know those public data sources are not considered authoritative by Google. On the other hand, it appears that even the authoritative sources they are using may not be “authoritative!
As you may remember, I recently suggested that Google better change gears and get a spokesperson who can speak cartographese (virtuous cartography), rather than mumble cartosis (silly replies about the map making process like the one above).
Of course, the trip to Newport Beach renewed my love of field research and I decided to continue my effort and see how accurately the new Google-base-US (GBU) had mapped my neighborhood.
Well, things seemed to have changed in my neighborhood since the last time I looked (at least according to Google), so I got out the Geotagger One and field checked the Google Map. Yep. I drove the streets, read the street signs and stopped at that – except for a couple of obvious geocoding problems. Below are a series of screen shots that show what I found (and did not find).
By the way, I zoomed Google’s map to its highest level of detail to determine whether a street was unnamed or if the name was not showing only at lower “magnifications”. Where I have shown a street name in red in the illustrations, Google did not provide a street name for that street segment. In other cases, I have drawn a red line through the Google street name when it was incorrect and placed the correct street name (as indicated from the street signs in the community) somewhere near the street.
Hope you don’t live on Poinsettia or Carlson, because Google won’t find you.
Also hope you do not live on Rolling HillS, Fair Court or Elder Brook. Golden Rod is a cul-de-sac that is unnamed on Google, although they have incorrectly renamed the rest of Magnolia as Golden Rod.
Camino Capistrano is not located on the left-side of the image, but, not to worry, is correctly located by Google at the right-hand edge of the image. In fact, the Camino Capistrano shown next Cabot Road runs along the edge of a flood control basin. The dirt path Google has improperly named Camino Capistrano (next to Cabot) is gated at both ends and is an access road for maintenance trucks, even though Google shows the path having an address range.
As you can see from this photograph, it would be hard to confuse the real Camino Capistrano will the one Google placed in the flood basin. Perhaps the erroneous location was based on using some of that “public” data.
Whoops. Seems like Google (or is it Garble?) has some problems with geocoding. I am sure the owners of Luisa’s Café know that success in the restaurant business equates with location, location, location and food quality. I suspect that is why their café is located on Cabot rather than the back edge of the warehouses, as shown by Google.
If you look at the first image showing Camino Capistrano, you will notice the shop for the mechanic who takes car of my car is located in the flood control basin. I guess this indicates either that Google is not using parcel maps for address geocoding or that the parcel maps they are using may not be “authoritative”. (You do remember my 6 part blog on “Authority in Mapping” don’t you? Well, it had 6 parts because the notion of “authoritative” sources is very important.)
Clearly my little test tells us nothing specific about the quality of Google’s coverage across the United States, but I believe that the type and number of errors described above are directional predictors of what Google users can expect to find when using Google maps at any location in the United States.
I compared Google’s mapping effort in my neighborhood with Bing Maps (database supplied by NAVTEQ) and found that the NAVTEQ map did not exhibit the problems I found in the Google map, although NAVTEQ missed one tricky road name change that Google caught. If Google built the Google-base to remedy the accuracy problems they were experiencing with NAVTEQ and TeleAtlas maps, they seem to have missed the mark, at least on their first attempt.
I think the interesting question, as I have stated before, is how will Google update the Google-base? While my own research sample shown above indicates that crowd sourced data could correct many of their mapping blunders, who might want to take the time to make these corrections? If Google cannot get these corrections from crowd sourcing, can they get them from map data in the public domain? Will either of these sources prove to be authoritative? Next time I will start with my thoughts on the updating conundrum facing Google.
By the way, Kevin Dennehy who writes on LBS for GPS World asked for my comments on Google’s purchase of NIM (Networks In Motion). You can find my comments in his article titled “Did Google’s Market Grab Spur TCS’ Purchase of NIM?”