Scrum is a methodology for iterative and incremental software development. Engineering practices, however, must also be adopted to help ensure the delivery of quality software. This paper relates the experiences of three teams at Microsoft who employed Scrum with nine engineering practices. Improvements in estimation accuracy, quality, and productivity are reported.

The evidence for improvements in estimation accuracy comes from team member views following their use of the planning poker technique. Regarding quality improvements, figure 3 clearly ranks defect densities for two of the Microsoft teams as better than benchmark data. The third Microsoft team, with a high defect density, had not adopted an 80 percent test code coverage criteria, and its tests achieved only 53 percent code coverage.

Many readers will be skeptical of the claim of a 250 percent productivity improvement. Figure 2 shows cumulative lines of code (LOC) produced by Team A. After four iterations, representing ten weeks of work, some 22,000 LOC were written by four developers. This yields a productivity of 110 LOC per day, more than double typical productivity rates from benchmark data. However, figure 2 suggests little or no increase beyond 22,000 LOC ten weeks further into the project. Productivity comparisons should be based on comparable project intervals.

Even though the productivity claim seems exaggerated, this paper does provide useful insights into the results of using engineering practices when applying Scrum. As such, it is strongly recommended to Scrum practitioners and those researching software development processes.