Data and our understanding of it can often be impacted by statistical outliers. These anomalies can significantly alter the perceived performance of a team by inflating averages, when in reality the performance is generally much healthier.

The outliers slider can be found in metric settings on the following key rate metrics:

Take the simple example of a team who delivered 5 stories last week. 4 of those stories took 4 days to deliver, but 1 of those stories was quite complex and required a number of iterations and weeks to complete. The team finally got it over the line, but in the end it took them 23 days to deliver the story. So whilst the far majority of what they delivered only took 4 days, the average time was 7.8 days. It's a crude example, but most teams have run into this issue when using raw averages to measure performance.

With Outliers, you can now remove these statistical outliers from your data in order to gain a truer view of your team's performance under "normal" circumstances. It's still useful to understand the full picture without excluding outliers, but having both at hand will make for a far more meaningful understanding of data, especially when communicating to external audiences.

Configuring Outliers

You will find the Outliersrange setting available within the metrics settings of our key rate metrics: Delivery Time, Flow Efficiency, and Code cycle time. You can adjust the slider (as shown below) to define only the percentiles on which you'd like to focus. By default, moving P0 to the right will exclude the lowest values, whilst moving P100 to the left will exclude the highest values. So if you were looking to measure the 85th percentile of delivery time, you'd set P0 to P85 (as shown below).

Delivery time: without outliers excluded (default)

Delivery time: 85th percentile

Code cycle time: middle 60th percentile (bell curve analysis)

Did this answer your question?