How to improve the hit rate when using Virtual Earth Platform Customer Services to geocode addresses

As you might know I wrote an article about Geocoding or How to transform a list of addresses into Latitude/Longitude values and display them on a map (including source code) a couple of month ago. The source code includes an Access database which holds the data to be converted. This database is later exported to an XML file which is then uploaded to the Virtual Earth Platform Customer Services to be geocoded and that’s it.

I recently received a mail claiming that the Virtual Earth Platform Customer Services used in the article lead to poor geocoding results.
As it turned out the problem was caused by the customers file formatting. The XML file must contain an AddressLine column which was misspelled (AdressLine).
Additionally the batch geocoding service is not very good with understanding localized country names (it is build on our legacy platform that we are planning to replace in near future). After he changed the CountryRegion names in his input file to English names (i.e. Germany, Italy, Switzerland) the success rate increased significantly 🙂

Batch Geocoding Results:
Entities geocoded: 478
Entities skipped: 0
Entities with good results: 404
Entities with imprecise results: 53
Entities with ambiguous results: 21
Entities with no results: 0