This paper presents a new approach to bottom-up stochastic occupant behaviour modelling for predicting the use of household electrical appliances in domestic buildings. Three metrics relating to appliance occupant behaviours are defined: the number of switch-on events per day, the switch-on times and the duration of each appliance usage. The metrics were calculated for 1,076 appliances in 225 households from the UK Government’s Household Electricity Survey carried out in 2010–2011. The analysis shows that occupant behaviour varies substantially between households, across appliance types and over time. The new modelling approach improves on previous approaches by using a three-step process where the three-appliance occupant-behaviour metrics are simulated respectively using stochastic processes to capture daily variations in appliance occupant behaviour. It uses probability and cumulative density functions based on individual households and appliances which are shown to have advantages for modelling the variations in appliance occupant behaviours.