The Importance of Data Quality in Software Reliability Modeling

Data quality is important to software reliability predictions, but in practice, engineers may need to make predictions using data collected during existing test activities. We faced this challenge for a large spacecraft ground control system. We were able to use software reliability analysis to build a case for allocating additional system test time, but our predictions for the amount of time needed were very inaccurate due to differences between the data that was collected and the data we needed. This paper describes the process we followed, the hurdles we encountered, and the lessons we have learned about the importance of data quality to performing accurate analyses.