Put it to the Test: Using Lightweight Experiments to Improve Team Processes

(Experience report, 30 min) [ Slides ]

Author(s): Michael Keeling (Institute for Software Research Carnegie Mellon University)
Session: Learning and Improvement
Session chair: Tore Dybå (SINTEF)
Date: Thursday, 3 June 2010: Main Conference
Time: 13:30-15:00
Experimentation is one way to gain insight into how pro- cesses perform for a team, but industry teams rarely do experiments, fearing that such educational excursions will incur extra costs and cause schedule overruns. When facing a stalemate concerning the use of pair programming one industry-like, academic team constructing a commercial-grade web application, performed a lightweight experiment comparing pair programming and programming alone using Fagan in- spection. Through the experiment, the team learned that pair program- ming was not only faster than programming alone, but also required less effort and produced code of more predictable quality. Conducting the experiment required only eight hours of effort over six weeks (a mere 0.5% of the total effort during that time frame) and afforded crucial in- formation for choosing the best practices for the team. As demonstrated by this experience, lightweight experimentation is cost effective and does not threaten project schedules.

<< Back to the program