How can we be sure that sufficient safeguards are in place and safety level is acceptable? As we heard Prof. Nancy Leveson stating at last year's MKOPSC symposium, even with all components functioning, dysfunctional component interaction can still be a cause of mishap. Human factor expert, Prof. Erik Hollnagel, asserts it in even stronger terms: the Efficiency-Thoroughness Trade-off principle, or rather dilemma, contends that one can hardly do it perfectly well. Perfect thoroughness, certainly in complex situations, requires an amount of time with which efficiency will be in conflict. For improved situational awareness, sufficient resilience, and adequate risk control, we must adopt a top-down system approach. Hazard scenarios possible in the system, with all its entangled interactions of hardware, procedures, and humans shall be identified bottom-up and causal relations made clear. Fortunately, in recent years two potentially helpful tools have become available: Blended Hazid, a vastly improved, heavily computerized system approach making use of HazOp and FMEA, and Bayesian networks, a tool to model cause effect structures allowing inclusion of uncertainty information. Bayesian networks as an infrastructure enable also the use of indicator values to relate the result of safety management effectiveness, which expresses itself as safety attitude of employees, competence, workload, and motivation, with their effects on error and failure probability. This paper will explain the directions these developments are advancing and the openings they provide for further process safety research and risk assessment, which when applied will result in improved process risk control. (c) 2014 Elsevier Ltd. All rights reserved.