We have a project which has approximately 2500 tests which are run in an average build.
For a given run, between 0 and 3 tests may fail (on average, 0).
The default view "Test Result Trend" shows a blue graph of all of our passing tests – the scale of the failing tests is invisible.
This makes the default view absolutely useless.
As it happens, the number of tests we run in a given run actually fluctuates significantly (50 one run, 1900 the next).
To make the graphs useful for us, a couple of presentations could be considered:
- Just show two graphs side by side. – Personally, I don't think this is a great approach.
- Use independent Y axes (typically a left axis for the total tests, and a right axis for failed tests) – this is fairly common in charts with mismatched scales. – In this model, the left y-axis label could be tests and the right y-axis label could be failures.
- Use an inverted graph for failures and stack them so that they share a common x-axis but total tests counts grow up, and failed tests grow down from the common x-axis (with independently labeled y-axes on the left).
- The label for the link to change the view show test # and failure # isn't really a great link text, it's both fairly long and annoyingly terse at the same time.
- The x-axis is pretty dense:
- It would probably be better show fewer build numbers.
- It would be better if there was more spacing between letters in the build numbers (I don't see any reason for the x-axis labels to be so constrained).