Refineries are highly complex installations and a potential source of major hazards. Due to the large volumes of flammable and toxic substances present, an accident in a refinery may have multidimensional consequences. This includes severe property damages, injuries to personnel, toxic releases of chemicals causing adverse health effects on nearby residents and the environment, and large business interruption losses that may lead to company bankruptcy. This paper looks at the risk profile of refineries from an insurers’ perspective. A top down approach is employed to derive key performance indicators (KPIs) for two types of events historically known as main causes of major accidents in refineries, i.e. fire and vapor cloud explosion. Bayesian Belief Networks (BBNs) are used to develop a probabilistic model for quantifying risk indication of refineries for fire and explosion events via a structured approach to elicit and synthesize available knowledge from domain experts. Three types of KPIs are modelled as BBN nodes: quantitative, qualitative and directional indicators linked to technical, human and change trend factors, respectively. The approach proposed has a twofold practical use: i) to support insurers to assess which plants have low potential risk exposures; and ii) to inform the refineries about their own risk profile, thus supporting them with the assessment and the implementation of risk reduction measures. To ensure applicability across the industry, the systematic development of the BBN is detailed and extension via the inclusion of modules accounting for further KPIs is discussed.