Using the lines of code (LOC) metric in software project management can be a financial moral hazard to an organization. This is especially true for upper management who handles an organizational budget and strategic plan. Software project managers have their own budgets. However, if they fail to meet the budget, the organization's cash flow, rather than the project manager's personal cash flow, will suffer. This chapter will discuss the practice of software project management, the field of software metrics, game theory, and the game theory issue of moral hazard. It will demonstrate why using LOC as a metric can present a moral hazard to senior management and an organization.
TopIntroduction
Cost overruns are significant problems encountered by the software development industry. One example is the United States Federal Bureau of Investigation’s (FBI) Sentinel software development project. As reported by the Cato Institute in Washington, DC (DeHaven, 2010):
Sentinel is currently 32% over-budget and two years behind schedule. Worse, an independent assessment of the project concluded that the project will need another $351 to $453 million and won’t be finished for another 6 to 8 years. As the inspector general notes,“… the longer the full implementation of Sentinel takes, the more likely it is that already implemented hardware and software features will become obsolete.
In 2011, a report from the Office of the Under Secretary of Defense, Acquisition, Technology, and Logistics indicated that after studying nine ERP systems, “ERP schedule delays and cost overruns were common and sometimes large” (G. Bliss, personal memorandum, February 23, 2011).
Those working in the software development industry can identify others within their organization or similar organizations.
A major cause of cost overruns in software development is the use of unsound cost forecasting methods. Often, software cost estimations are poorly taught or overlooked in IT academic programs. Operations research/analytics programs rarely address the issue or will teach unsound methods. Probably less than 10 successful software cost estimation management consulting firms operate in the U.S.
An effective approach to forecasting software development costs is to look for a strong statistical regression correlation between a predictor variable (representing the amount of software developed) and the resulting cost to develop the software. While many predictor variables have appeared in practice, the LOC variable has a strong following in both government and industry. Written in a programming language, programmers generate LOC. According to the LOC metric approach, as software requirements increase, more LOC will be required. Work to write the LOC (as well as perform testing, deployment, and project management tasks) will increase as the number of LOC increase.
This chapter will detail the LOC metric and show how fundamentally in error its use can be.