How we calculate flow metrics

ℹ️ This feature is currently in our beta phase. This means we are testing it with our customers. The user experience may also not be ideal yet. If you are running into issues, please let us know at support@columinity.com

On this page, we explain how flow metrics like cycle time, throughput and work-in-progress are calculated in Columinity.

Cycle Time

Cycle Time is "the average time it takes to complete one item, from start to finish". It indicates how long it typically takes to complete a work item, and can be used for forecasting. We calculate this metrics according to the Kanban definition:

1. Collect all work items that (first) started and (last) completed in the past 28 days. We consider an item "in progress" when it leaves the "To Do" state, and "done" when it hits the "Done" state.
2. For each work item, calculate the number of days between start and completion.
3. Calculate the time (in days) it takes to finish 85% of the work items. This is the SLA 85, or 85th percentile of the Service Level Agreement (SLA).
4. Calculate a 95% confidence interval with a 3.000 bootstrap sample. The narrower the confidence interval centers around the SLA 85 result, the more reliable it is for forecasting.

Work-in-Progress

Work-in-Progress (WIP) is "The time-weighted average number of work items in progress during the measured period". It indicates how many work items are typically in progress at the same time. Kanban shows that a lower WIP is generally better, as higher WIP can indicate dependency issues, low collaboration and lacking cross-functionality. We calculate this metrics according to the Kanban definition:

1. Collect all work items that (first) started in the past 28 days. Uncompleted items are included. We consider an item "in progress" when it leaves the "To Do" state, and "done" when it hits the "Done" state.
2. For each day in the sampling window, calculate how many items were "in progress".
3. Calculate a mean average work-in-progress.
4. Calculate a 15-85% confidence interval for the WIP result. The narrower this centers around the WIP result, the more reliable it is.

Throughput

Throughput is "The average rate of completed work items per unit of time". It indicates how much work a team typically completes every day. We calculate this metrics according to the Kanban definition:

1. Collect all work items that (first) started and (last) completed in the past 28 days. We consider an item "in progress" when it leaves the "To Do" state, and "done" when it hits the "Done" state.
2. For each day in the sampling window, calculate how many items completed on that day.
3. Calculate a mean average for the items completed per day.
4. Calculates a 95% confidence interval based on the standard deviation.

Little's Law

Cycle Time, Throughput and Work-in-Progress are connected through Little's Law. This means that one can always be calculated from the other two under a few conditions:

WIP = Throughput × Cycle Time
Cycle Time = WIP ÷ Throughput
Throughput = WIP ÷ Cycle Time

Little's Law applies when the system is in a steady state, meaning that all items started and completed in the sampling window. Moreover, the Cycle Time should be calculated as the mean average of item cycle time's, and not the SLA 85. We validated our calculations in Columinity by ensuring that, under the aforementioned conditions, the law applies to the returned cycle time (mean average), WIP and throughput.

See issues?

We worked hard to ensure that the calculations above are as correct and reliable as we can make them. If you see areas of improvement, please contact us at info@columinity.com

Related resources

• How to configure and use the Azure DevOps Connector
• How to configure and use the JIRA Connector