Anyone who has had anything to do with failure rates of industrial equipment will not fail to recognize that this makes Pro-M predictions essentially indistinguishable from a generic source of reliability data, at least for the component types examined in this study.  That this is already a remarkable achievement can be appreciated if you run your eye along the horizontal axis and observe the drammatically different component types and the enormous range of complexity spanned by Pro-M in tracking the failure rate so accurately between them.

The error bars shown in the chart do not indicate the accuracy of the data, nor are they meant to indicate the accuracy of Pro-M predictions.  They are there simply to show the observer what a factor of 4 either side of the Pro-M prediction looks like on the chart.  Although the error bars are attached to the Pro-M points there is good reason to believe that much of this uncertainty can be attributed to the lack of knowledge of the preventive maintenance that was being carried out on the sample equipment.  Common to most generic databases, the EIReDA database provides only very sketchy and incomplete information about the applied PM tasks and intervals.  Consequently, the Pro-M predictions, which are strongly dependent on this information, would better be displayed as a band covering a range of values representative of the possible PM assumptions that could have been made.  In that case there would be no daylight between the range predicted and the generic values.

There are also independent theoretical reasons to suggest that Pro-M predictions for a given component type will be rather more accurate when addressing changes in the relative values for the same component type as a function of in-service stressors, duty cycle and PM tasks and intervals, because various sources of uncertainty that come into play between different component types will not be present, while others will tend to cancel out.

Pro-M's ability to estimate failure rates for complex, high-end industrial equipment as a function of operating and in-service conditions has resulted in a rapidly increasing usage of the database for predicting hard-to-find failure rates for certain component types involved in life cycle maintenance and long term asset management studies.  Pro-M should therefore be considered for this purpose, and also for reliability and maintainability optimization in engineering design applications.