IFPUG Announces the Publication of ‘The Statistically Unreliable Nature of Lines of Code’ in Crosstalk Magazine

June 5, 2005 Uncategorized

PRINCETON JUNCTION, NJ, June 5, 2005 – The International Function Point Users Group (IFPUG) recently announced the publication of the article ‘The Statistically Unreliable Nature of Lines of Code’ by Joe Schofield in April 2005 Crosstalk. According to Mr. Schofield, “This article suggests that the venerable line of code measure is a major factor in poorly scoped and managed projects because it is itself a vague, ambiguous, and unsuitable parameter for sizing software projects.” The data for this important article came from several Personal Software ProcessSM courses taught at Sandia National Laboratories in New Mexico. The demonstrated variation in LOC counts for programs that were written to the same specification (all programs were verified to have the same function point count), in the same language, verified by the same certified instructor, ranged from 2200 percent in the worst cases to 300 percent in the best cases. The 500+ data points came from 60+ attendees. Mary Bradley, President of IFPUG indicated, “Mr. Schofield’s article is an important addition to the software measurement body of knowledge. Anyone involved in measurement or estimation should read this article and understand the ramifications of using unreliable data.”