This paper describes work recently undertaken by scientists and engineers at Arup Fire in Leeds, University of Manchester, London Fire and Emergency Planning Authority and Southbank Univeristy into the stochastic aspects of fire in shops and commercial buildings. The work is based on data from the 'Real Fires Database' which includes more detailed data on fires than has traditionally been collected. New data on the number of people, fire growth rates, detection times and pre-movement times is presented and discussed. The potential for dependencies in and the limitations of the data are also discussed.