We basically use these three workflows (or slight variants of it) when creating new rules.
To compose a rule
If we want to compose a new rule, we create a .TXT document on the Desktop, put there a couple or three of the instances as sourcetarget. We then save the .TXT file, right-click on it and choose Run QA in Xbench to launch an Xbench project with just that file.
Then we use the Search tab to compose the search. Once the search does what we want, we click Add last search to checklist.
To test the effectiveness of a rule:
We have a large existing Xbench project with many items (in the range of hundreds of thousands of segments). Once we have added the checklist item to the checklist, we go to the Checklist Manager, select the entry, right-click and then choose Test. It allows us to see how the rule behaves with a larger data set.
To test the performance of a rule
With the large existing Xbench project, we run a QA with the full checklist where the new entry is found. Then we go back to Checklist Manager and look at the Run time (ms) column to see how the new entry compares to existing entries.