During a playback, Testimony is looking for differences between outputs seen in the recording and outputs seen in the playback. For example, in a dialog transaction it will be looking for unexpected messages (a message seen in the playback that was not seen in the recording) or an unexpected next screen (where the playback navigated to a different screen than was seen in the recording). Whenever a different output is detected during the playback, Testimony marks the step as having failed, and the script as either completed (if the failure occurs on the last step) or partially complete (if the failure occurs on an earlier step).
Once the playback is completed, a defect proposal is run which groups identical failures into defects. For example, if a transaction has failed 25 times with the same error in the playback, then all 25 occurrences of this failure will be grouped into one defect.
The defect record contains lots of useful information about the defect which can be used to aid diagnosis and perform defect investigation and management functions. These are discussed in the following sections.