Emily presents an overview of different feature location techniques (FLTs) and asks the question: which of them are better and how can we compare them? Current comparisons are a bit unfair, as only the precision is compared. That means a method that first lists 4 unrelated pieces of code and then the bug is just as good as a method that results in 4 related pieces of code and then the bug.
The new method that Emily proposes is to compare the likelihood of finding the bug. So if related pieces of code are in the top of the list, we assume the developer will inspect them and will find the bug earlier, if the parts of the code are more related.
This figure gives an overview of the results. The results demonstrate that traditional IR measures such as precision and recall can give a misleading picture of the effectiveness of a feature location technique, when a technique can produce a ranked list that allows the developer to navigate to a ﬁx location quicker.