Determining Software Product Release Readiness by the Change-Error Correlation Function: On the Importance of the Change-Error Time Lag
In software development determining the release readiness plays an essential role. The number of errors is frequently used as an important measure to decide about the quality of a software implementation. Therefore, error prediction techniques have been intensively studied in the literature for many years. Despite this, their adoption in practice is still strongly limited to date. In this paper, an alternative model for error prediction in software projects based on linear response theory and the change-error cross correlation function is proposed. It is applied to data collected in projects of a major embedded systems vendor in the communication industry. Under similar conditions, a universal behavior of the change-error cross-correlation function is observed. Moreover, a time lag of 4-6 weeks between the change and the detection of related errors is discovered. This clearly demonstrates that for reliable release decisions not only the current number of errors but also of changes is essential.