We’ve previously talked about our Burn Confidence band, which measures team performance as compared to expectations in a static representation. However, the development lifecycle is constantly changing and requires teams to respond dynamically. We are excited to announce that our latest update to the Burn Confidence Band transforms the band into a dynamic forecasting tool based on a rolling sequence of burn data from past sprints.
The new band aggregates the last 16 sprints of burn data. The band is constructed from the ‘Work Done’ lines from those 16 sprints. To make sure that the band represents an average burn up pace, we remove any outliers from the dataset. This means that for a given day in the sprint, the band is determined by using the maximum and minimum burn values for that day from those previous 16 sprints. In layman’s terms, you could say “On day 8 of the sprint, our team usually burns between 4 and 20 points”. We form the band using those values, recalculating on each day of the sprint.
We decided to remove outliers from the dataset before calculating the band values. This prevents the band from becoming too wide because of exceptionally good or bad sprints. More importantly, a narrower band provides a more precise representation performance so that leaders can confidently articulate how teams are tracking.
We remove outliers by filtering the dataset using a 1.5 standard deviation threshold. This filters out approximately 13.36% of data from our dataset (approximately 6.68% from the high end- ie: good sprints, and 6.68% of data from the low end- ie: bad sprints). We can visualize this using a bell curve where the data outside the red area is filtered out as “outlier data.
Simply put, we retain 86.64% of the burn data by filtering any data outside 1.5 standard deviations from the mean.
After the release, the development team noticed the new Confidence Band could help forecast the likelihood of a ‘clean sprint’ at a glance.
Take the above chart as an example. We can quickly see that this team rarely burns more than 50 points by the end of this sprint, even in the best possible cases. However, they were initially assigned a goal of 62 points for the sprint. Looking at other sprints from the same customer we could see this pattern repeatedly. This sets the team up for failure, which is bad for morale and performance.
How can this new band help with sprint planning? By using the band as a visual aid, we can easily estimate the number of points a team is likely to burn by the end of the sprint. This can help build a sprint plan as follows: