We had a long discussion about what went wrong and how to do better. One thing was that although we identified 6 "steel threads" (aka thin slices, tracer bullets) in our big UI story, we didn't finish the third thread until the 7th day of the sprint - at which time most of four more threads were also done. It was incredibly confusing to try to test it, we couldn't tell what should work and what shouldn't, and we missed a couple of gaps in the features. Plus, unit tests were being written after the fact.
Here's our stop/start/continue list that resulted from our retrospective:
- Transfer whiteboard drawings from conference room to Scrum arena, and check off threads as they are done
- First thread of first story has to be done by first Tuesday of Sprint
- Log remaining hours on task card (in Mingle) prior to Scrum
- Test first
- Working more than one thread ahead of unfinished thread