Research Case Study

Impact: Imagine being able to predict a crime in the future

We investigated whether Milan’s predictive policing strategy really gets results.

  • Tagged under

    Economy, business, politics and society
    Global perspectives and challenges
    Technology, data and innovation

  • Lead Academic

    Professor Giovanni Mastrobuoni

Photo of an Italian police vehicle

Advanced computer software designed to enable the police to predict both the perpetrator and the timing of a crime, saved one Italian force close to 2.5 million Euros in one year.

Those were the findings of the programme’s first academic evaluation carried out by Professor Giovanni Mastrobuoni from our Department of Economics

The software programme is called KeyCrime and is used by Milan Police Department, the Polizia.

“When this advanced yet inexpensive IT innovation is used, differences in police productivity are striking,” explained Professor Giovanni Mastrobuoni, from the Department of Economics.

Read Professor Mastrobuoni's full research paper on the Essex Research Repository

How it works

Robbers are creatures of habit and criminal gangs tend to select the same business types, around the same time of day and in the same city or neighbourhood, especially if previous robberies have been lucrative.

"When these habits are properly tracked and identified, that predictability can be put to effective use" 
 Professor Mastrobuoni Department of Economics, University of Essex

The software collects and analyses around 11,000 bits of information about each robbery (time, date, location, type of business, type of crime), about the criminals involved (perceived age, height, body structure, skin, hair, eye colour, clothing), about the observed weapons (type, make, model, colour) and about the observed vehicle (type, make, model, license plate).

This is then combined with police reports, interviews with the victims and surveillance camera footage before using comparisons to establish links in order to identify and predict criminal strategies.

Professor Mastrobuoni added: “Most re-offending occurs within a few days, which means that at any given point in time there is a limited set of unique groups of robbers whose actions need to be predicted by the software. When all of this information then becomes available to patrols out on the streets, it puts the police in the right place at the right time. There is no doubt that this type of micro predictive policing is a highly effective, efficient IT Investment.”

Professor Mastrobuoni is currently working with Essex Police in developing and evaluating effective predictive policing practices.