您在這裡

How much testing is enough?

20 一月, 2016 - 15:30

A central question is, when are you done testing? We have already said that it is impossible to guarantee that a system is free of defects, no matter how long you test it. So when do you stop?

This is essentially an economic question. It can be formulated in two ways:

  • The benefits of delivering the system, including its residual errors, outweighs the risks, or
  • The cost of discovering additional defects is greater than the savings gained from finding and removing them.

Either case boils down to a judgment call. This is most apparent in the first formulation, which refers explicitly to risk. In the second view, we need to add the word “probable” in front of both the cost and the savings, to emphasize that it is hard to predict how much effort you will need to find the next bug and indeed how grave that defect is going to be.

The decision is illustrated by the defect removal curve in Figure 3.8.

media/image10.png
Figure 3.10 Defect removal curve