Abstract: A method is provided for updating a fault tree used in diagnosis of a condition of a machine. The fault tree is a sequence of test modules. Each of the test modules is assigned a confidence score indicating the likelihood that the test module will lead to a diagnosis. The method comprising the steps of obtaining service data from a plurality of service occasions for like machines, revising the confidence score for at least one test module in the fault tree based on the service data; and revising the sequence of the test modules in the fault tree based on the revised confidence score(s). The sequence is revised such that the most likely modules to result in a diagnosis are listed first, and the least likely ones are listed last. The re-ordering is based on feedback from actual service instances in the field, and thus is more likely to reflect real-world conditions. The result is that service technicians using the revised fault trees work more efficiently, at least on a statistical basis.