This is another post relating to the notes I took at the 3rd Toronto Workshop in Software Testing.

In addition to providing an excellent space for the event, Jon talked a bit about how they try to quantify risk in his employer’s bug system. In addition to a number of other criteria, they throw ‘Frequency’, ‘Distribution’ and ‘Productivity Impact’ into a calculation which determines the fix order of bugs. Each of the three categories is presented as a drop-down menu with easy to understand values, Frequency has ‘Once a day’ for instance.

The big thing for me was that the choice of appropriate labels is perhaps more important than their input to the calculation. There was another field (which I neglected to write down) which was going to be removed because no one used it. And why didn’t they? It wasn’t because the information wasn’t valuable, but that the labels were being interpreted differently by different people and so lost it’s usefulness.

Two other random points I have written down are:

  • Optimize your bug system for entry to encourage people across the company (not just in test) to log bugs
  • Severity is often only in the context of the code, not to the customer