An interesting debate arose the other day within a planning session we were having – how best to estimate what we could get into the current sprint.
The team had little idea of its velocity due to story points either changing after review or stories coming in mid-sprint with no points assigned. Large numbers of stories had been consistently carrying over and task estimates were poor. Given that the task estimates were poor, to me there seemed to be little point in planning using them, so I backed the use of story points. There was mixed opinion in the team.
In having a look around the available Agile literature, I stumbled across an article by Mike Cohn which backed the use of task-based hours planning. The article made sense to me on the grounds that you have more knowledge at the start of the sprint to get into more detail. Furthermore, to me, task planning also forces you to think about what you need to do to accomplish the task in hand.
However, what if your velocity suggests you are working at 15 story points a sprint but your task analysis suggests that you can do 40?
A middle of the road way forward would be to use task based estimates but cap the number of stories to within (say) 20% of the current velocity. Stories can always be brought in mid-sprint if the team agrees (maybe at stand up) that there is enough capacity to support the new story coming in.