Modern process computers of petrochemical plants continuously acquire and store huge amounts of data which can be used for process improvement. As an example the problem of hot spots in reactors of a Shell petrochemical process was analysed. It is explained how data acquisition and statistical data analysis can be used to extract information about reactor performance and aid in problem analysis. This analysis of plant data with PCA (Principal Component Analysis) and cross-correlation techniques shows that maldistribution in the reactors is the primary cause of hot spots. Solutions, based on improved catalyst loading techniques, have been proposed and implemented. They yield about 2 million dollars extra each year. We show that an in depth analysis reveals causes and solutions to problems in operation of plants, which have been unsolved for many years.