Each year the CSB provide the process safety community with reports of past accidents in the US process industries. These reports often contain a detailed analysis of the event and its consequences. Often the report finish with recommendations to many stakeholders. This paper analyze the causes and recommendations in CSB investigative reports from the period 1998 to 2011, by looking at questions such as: How many recommendations does a report contain? Which organizations are they directed at? What systems are most frequently parts of causes or recommendations? And then attempts to look at the question: Are the recommendations directed at the right organizations? - if the goal is to improve process safety. How many recommendations should a report contain to have maximum impact on the learning form the event? What changes have resulted from past recommendations? How can the learning from the investigation reports be improved?
