[Previous entry: "JUnit Recipes: Great book, less Maven..."] [Next entry: "Contract Tests"]

02/28/2005: "Predictable Delivery: A killer metric"

A killer metric. So many projects collect so much information about the day-to-day goings-on, while many other projects collect next to none. Are you overwhelmed by the sheer volume of measurements you could be taking and recording that you choose to do none? Would you rather track one killer metric that correlates highly with your team's ability to deliver predictably?

Mean time between injection and detection. Measure that: the time it takes to detect a new defect once it has been injected into the system. Not reported, mind you, but injected. Which lines of code were changed to fix a defect? When was that code last changed? That's the latest time that defect was likely to be injected. Certainly, a lower bound measurement is adequate for our purposes.

The higher the mean time between injection and detection, the higher the cost of the defect. Think about what your team goes through to fix a defect, including describing it, logging it, recreating it, arguing over it, prioritizing it in meetings, tracking its root cause, fixing it, testing the fix, committing the fix, then verifying the fix. When we detect the defect weeks after it is injected, all these costs are substantial. When we detect the defect seconds or minutes after it is injected, many of these costs are eliminated.

So, find defects moments after they are injected. Sound impossible? Ask us and we'll show you how.