More on the Mysteries of Geocoding
A couple of weeks ago, I wrote a couple of articles about common geocoding errors. Last week I was at Where 2.0 and spoke to some of the companies that had supplied data to the mapping efforts that were the focus of my articles.
Note: I did not tell these people that I had blogged about geocoding and that they had failed the test. Instead, I asked them about their quality standards, algorithms and plans for future improvement. I was stopped in my tracks when I asked someone from Microsoft about their use of geocoding. The reply was that they now used “roof-top” geocoding. I pursued the opening and said, “You mean you are using parcel data and building identification data from companies like First American Flood Services or Group 1 software?” They replied “Group 1”. I asked “If I go to Windows Live Local and search for businesses, then the location of the symbols I see will be based on Group 1 data?” The answer was “Yes.”
That led to consider that SuperPages, which was featured in one of my blogs, although using Microsoft Virtual Earth for Mapping might not be using the Microsoft product for geocoding. A second possibility was that SuperPages was using a different source of listings data than Microsoft. So, when I got home I took a look.
I remapped one of the areas I wrote about last week using SuperPages again, but I then mapped the same area using Microsoft Live Local. Both pages are badged with the tag – Microsoft Virtual Earth™.
SuperPages still had the same errors. However, the listings data appeared identical to the addresses that Microsoft was using in Live Local, so that was not the source of error. Instead, it appeared to confirm that SuperPages was passing lat/lon data to Microsoft, rather than supplying address data for the listings and asking that it be geocoded by Microsoft. The actual locations of the coffee shops are shown by the white numbers inside of black circles.
When I looked at the same coffee shops in Windows Live Local, the locations for the same two shops were clearly different from SuperPages, but still no cigar. The actual locations of the coffee shops are shown by the white numbers inside of black circles. It appears that the Group 1 data used by Microsoft does not included details for store locations within strip malls.
What does all this mean? From a surface view (which is all we have when we look at displays such as these) it appears that some providers of local search are making it very difficult for their search customers to close the deal. Of course, the use of non-harmonized maps and data also makes it impossible for searchers to find the shops of companies advertising with SuperPages!
It is difficult to understand why anyone would knowingly create a local search solution that was unable to accurately locate the destinations of interest to their customers. In SuperPages favor, the mapping is clearly labeled “Beta” – guess that makes it OK to have erroneous maps.
Next time, I will report on an interesting conversation I had with InfoUSA about their strategies for improving data quality in listings.