When Humberside police decided to re-engineer its criminal intelligence databases, which include information about offenders, victims and vehicles, it knew it had to get the data in good order first.

Not only was there a problem with people giving deliberately false information, there was also a lot of incomplete data, because someone reporting a crime, for example, might be too distressed to give full details of names and addresses. And the sort of people who end up on a police database are often not the sort of people who stay in the same address for very long, so the quality of data degrades very quickly.

For Humberside, the pressures have been greater than for most. In the Bichard enquiry into the Soham murders, the force was blamed for poor information management.

There were three aims in cleaning the data, says Graham Dawson, head of information services at Humberside Police: to ensure operational effectiveness; to have accurate intelligence information; and to allow Humberside to share accurate data with other forces and partner agencies. The ability of the police to share data with other agencies is a central part of government programmes such as the Cross-Regional Information Sharing Project.

If an officer is responding to a call about domestic violence, for example, they need to know there have been other call-outs to the same address. ‘Officers often say: “if I had known that bit of information, I would have dealt with that situation differently’’,’ says Dawson.

A key aim was to move to a set of golden records, where all the information related to a particular individual was held in the same record, rather than scattered across a series of records held in different systems.

Since last March, the force has been using Informatica Data Quality to check and clean the data. There were plenty of discrepancies. In the first 25,000 records looked at, there were 2,000 duplicates. Among other errors and missing data, nearly all the nominal records were missing the individual’s title, such as Mr or Mrs. Only 83 per cent of the vehicle registration numbers conformed to the agreed Humberside standard.

After a data cleansing and data quality profiling exercise, all the records now hold individuals’ correct titles and 98 per cent of the registration numbers conform to the standard.

The software has also enabled a team of intelligence analysts in the force to identify links between known individuals and crimes, and between incidents recorded in different systems. Data from the different systems is gathered in a data warehouse. The Humberside data warehouse and the South Yorkshire data warehouse can be searched together using an Autonomy search engine, benefiting both forces.

The battle against poor data is, acknowledges Dawson, a war of attrition. But he believes the police can now find accurate, relevant data more quickly than a year ago.