Brent initiated a crime mapping project with the Metropolitan Police. Tests conducted by the project showed that if a comprehensive address list of all properties in Brent was used as the geo-referencing database to match against crime records extracted directly from the Metropolitan Police’s crime reporting database, only 8% of those records would be successfully geocoded. These problems were due to poor cleanliness, structure and data management of crime records. On average, 3000 records required processing each month. This meant that approximately 2760 records required manual geocoding each month. Each record would take approximately 2 minutes to manually geocode, equivalent to 13 working days.

The solution that Brent adopted was to use an automated system that cleaned address information in the crime and disorder records, perform intelligent and logical geocoding routines against this clean data, and sanitise the records to enable information sharing. The solution:

• reduced the time taken to geocode one month’s of crime records from 13 working days to less than half a day
• reached a geocoding hit rate of 99% (i.e. only approximately 30 records each month required manual geocoding)
• and recorded geocoding accuracies at property precision of 90% and at postcode precision of 96%.