To improve safety one has to know where risks are. For determining risks, hazards have to be identified and representative accident scenarios defined. This needs effort and technique. Man is quite limited in foresight without having experience and lessons from the past. For knowing the risk of an incidental, undesired event both its severity and probability has to be estimated. Then ways to reduce risk become clear. in a process plant risks are many and it is not possible to remove them all. One has to attribute priorities. Intuitive and qualitative methods can do much, but plant complexity may be large and communication on risk may become difficult without formal methodology. Quantitative risk analysis offers much, but has its weaknesses and drawbacks. The required effort is considerable, specialists are needed, and variability in answers is large. Yet, a model built to go along with the life of an installation and updated periodically may be very useful. This paper presents an overview of the demand, problems encountered, possible remedies, and an outlook on useful improvement and extension of risk analysis methodology, including decision making. (C) 2009 Elsevier Ltd. All rights reserved.