There has been much work dedicated to crime analysis and intelligence in recent times. Independently, physical evidence has shown great potential for linking crimes and bringing solid informative data through the increased use of multiple databases. However, their informative potential is still often underestimated and has been poorly integrated into police information systems. We propose a framework that fully introduces this data into an intelligence based system. This framework is built on the study of inference structures extracted from investigators’ every day implicit reasoning processes. Five specific inferences are studied with the particular problem of serial burglary investigation across independent police and legal structures. On the basis of such an analytical approach, a computer prototype has been designed; it has shown great promise and has resulted in several operational successes.