Way back in my distant past there were all sorts of ways of managing and tracking software bugs. The general opinion of the developers and management was that we had to:

• find a specific number of bugs based on historical data (bugs per line of code)
• close a certain percentage of the more important bugs

The first point is specious. Bugs depend on all sorts of factors and ‘bugs per line of code’ is used because it’s easy to measure. Really though it’s a useless measure.

Having just released the first version of a complex web application ‘into the wild’, I’ve been pondering an alternative measure, which should work for all systems. Provided that you have an effective testing regime using real data, as time goes on the frequency of bug discovery will reduce. Furthermore the type of bugs found will become more trivial either in their consequence or in effort to fix.

If you track bug detection, normalised by severity and discovery effort this should provide a metric by which a product is then deemed ready to release. Just a thought.