I worked with
Parker Bannister to analyze the performance of the engineering org over H1 to date. We analyzed sprint-over-sprint velocity mean and deviations. (Yes, this should be a log-normal distribution. @ me if you have a good refresher course on log-normal analysis.) We also analyzed story cycle time versus point estimate.
Sprint Analysis
- Collected sprint data
- Calculated effective team capacity for each sprint due to vacation and holidays
- Calculated capacity-adjusted velocity
- Analyzed planned versus unplanned tickets
- Calculated mean, median and standard deviation
Story Estimate Analysis
- Collected story data for each team
- Calculated cycle time durations for each story
- For each estimate value (Fibonacci values 1, 2, 3, 5, 8, 13) cycle time quartiles were calculated
Results
We were able to make recommendations for each team based on results, primarily what point total each team should target in order to predictably complete assigned work. We also eliminated 3-pointers on the Data team due to the data showing that 3 and 5-point stories were statistically indistinguishable.