To Split or Not to Split, It’s Really Complicated.

At our most recent sprint review, a team I’m coaching completed all but three stories. So at sprint review, we were trying to decide what to do with the stories and whether to move them into planning or split them. All three stories had outside dependencies. All three had completed tasks and two of them were still blocked by the dependencies. While there could be a lot of second-guessing as to how we got ourselves into this fix, the burning question at the time was what to do about the points. We looked at three possible scenarios:

1. Moving the story to planning and all the points with it.
2. Splitting the story, leaving a 0 point story with finished tasks, and moving the story to planning with all its points.
3. Splitting the story and re-estimating each part of the split.

Now, I will say that this team is new to Agile and has recently started tacking Velocity Variance and Story Completion Ratio in an effort to understand what impacts their predictability. As with all new metrics, the team is very focused on them. We are calculating the Velocity Variance as Velocity this sprint – the three sprint average divided by the three sprint average times 100. Story completion ratio is calculated Stories completed from sprint planning/ stories committed at sprint planning. You can see the angst this could cause with three stories affecting both measures.

Scenario 1 – Move the story, all its tasks and all points to the next sprint. I have worked in teams that took this approach. We did not mind a few lumps in our velocity trend. For this team though, there was an objection to inflating the sprint by three stories, one of which was a 5 pointer. They saw the extra points as a potential ding for the Variance calculation. Also they had just planned a big sprint about two sprints back and didn’t make the commitment so were shy of having a really large number. Also, having the story with its completed tasks they felt was misleading, as some of them were rather large so the math just looked funny. Day one of the sprint I have completed 15 hours toward a story.

Scenario 2 – Split off the completed tasks into a 0 point story and plan the remaining tasks in a story with all the points. This solved the problem of having too many completed hours in the sprint but still caused the sprint to seem inflated. The other benefit of splitting was in our tracking tool there was less scrolling and the team could focus on the tasks left to complete. The velocity on average would continue to be as accurate as possible though the Variance could still take a hit on 2 sprints, the Story ratio would only be affected in the “failed” sprint.

Scenario 3 – Split off the completed tasks into a story, move the story with incomplete tasks to planning and re-estimate each part of the split. Although not the classic way splitting is taught, it would alleviate the concern over velocity variation. This prompted a lot of discussion on whether the split actually gave value and could be a “real” story indicating that the story itself might not be small enough. When the team ran a couple of examples we found that most of the incomplete tasks were either data manipulation or testing. The team felt this indicated that the stories had been held by the blocking issues and not too large to complete with data and testing in the sprint.

The team chose to implement Scenario 2 with the modification that they would not plan blocked stories in a sprint. In the case of the two blocked stories, one was waiting on a data file and the other was waiting on wording from the legal team. This lead to another best practice for our team, plan stories only if all of the external dependencies have been resolved before sprint planning.

The biggest win was the lively discussion and the empowerment the team demonstrated resolving their incomplete work issue.