Scrum Burndown, Remaining Effort and the Fibonacci Sequence
I'm a scrum practitioner and a Certified Scrum Master (I may soon decide to get my probably worthless but yet rewarding title of Certified Scrum Practitioner). I recently added some custom rules to Team Foundation Server work items related to the remaining work estimates. In my team(s) we always use a Fibonacci Sequence for our initial estimations during out estimation poker sessions. If you are unsure why many people tend to use a fibonacci sequence in estimation then Google it and I'm sure you'll find no shortage of explanations. Some of them might even be valid. I won't go into it here except to say that it helps address the concept that larger estimations are inherently less accurate. I put into the remaining work estimate the allowed values being the same fibonacci sequence. If you can't estimate it with more accuracy to begin with I don't think you can estimate it with smaller granularity now. Granted you are closer to the problem but no matter how close I am to the problem 21 hours can pretty easily look like 34. Maybe I'm just bad at estimating.
The problem I ran into is that though my team has taken well to the fibonacci sequence in the original estimates and we have great planning sessions of discussion and estimation poker - the daily remaining effort is still a hurdle for them to estimate in a fibonacci sequence. I felt pushback (not least of which from myself) at having worked at a task and not being ready to commit to 13 hours but really not wanting to leave it at the same 21 it was at when it was started the previous day. I fear people would too quickly jump down the one level when the remaining work number just felt that it shortchanged them their efforts and then we'd get stagnation at the 8 hour range where it would sit until the 'real' number caught up.
Granted the real solution is to estimate more granular tasks so that the jumps in the sequence are smaller and likely more accurate and then in reality the sequence becomes pretty much insignificant but... for now most the tasks seem to come in at 21 hours. About twice what I would like but in time we'll get more granular. We have seen our per capita velocity double in the last 6 months, so I am not going to knock it.
So... I relented on what my logical nature desired (keep all estimations with increasing magnitude) and followed my much more pragmatic gut (we should estimate daily remaining work freely) and allow the daily remaining work to be any value. I think it gives a more accurate feel to the person owning the task and they are more likely to say 'I worked on this a full day, but you know, I think it will still a few more hours than I thought yesterday even after my efforts to date'. My conclusion after experimenting briefly with remaining work as a fibonacci is that it may make the same technical sense but the human factor makes it a non-starter.
Do you estimate remaining work as an unconstrained integer or from a set of allowable values?