Abstract
Although software measurement is a valuable management tool, software metrics are often not as useful as practitioners hope. Using data from large corporate databases, the authors describe how the ISO/IEC 15393 standard gives inappropriate advice for measuring software engineering processes. They also show how this advice, when combined with the CMM/ CMMI level 4 requirement for statistical process control, encourages the use of misleading metrics and inappropriate data aggregation and analysis techniques. They summarize lessons learned and recommend using small, meaningful data sets and effort-estimation models to assess productivity. © 2007 IEEE.