metricDelivery & agility

Cycle Time

Cycle time measures elapsed time from the defined start of work on an item until that item reaches the workflow's defined finished point.

Download Markdown (.md)

01What it means

Cycle time is the elapsed time between a work item's explicitly defined start and finish points. It includes both active work and any waiting, review, handoff, or blocked time inside those boundaries. Teams must state the boundaries and calendar convention—such as elapsed or working days—because different choices produce different numbers.[1][2]

Elapsed time inside explicit workflow boundaries Cycle time includes active work and every wait between start and finish
Cycle time = finish timestamp − start timestamp
Start boundary
Observable point where work enters the measured process
Finish boundary
Observable point where work meets the defined completed state
Included time
Work, queue, review, handoff, block, and delay
Calendar
Elapsed or working time stated consistently

Report the distribution and percentiles, not only the mean; different item types, endpoints, and calendar conventions produce different measures.

02Why it matters

Cycle-time data shows how quickly and predictably work moves once started. Its distribution can reveal unusually old items, queues, batch behavior, and changes in delivery reliability. Percentiles are usually more informative than a single average because knowledge work varies and a few long items can materially affect expectations.

03When to use it

Use cycle time when improving a workflow, setting a service-level expectation, comparing delivery patterns before and after a process change, or answering how long similar started work usually takes. Segment the data when item types or classes of service follow meaningfully different paths.

04How it works

  1. Define observable start and finish events and apply them consistently to comparable work items.
  2. Record each item's timestamps and calculate elapsed duration using one agreed calendar convention.
  3. Plot completed items as a distribution or scatterplot rather than reporting only the mean.
  4. Inspect percentiles, trends, and outliers; connect them to queues, blockers, handoffs, or oversized work.
  5. Change the workflow cautiously and compare subsequent distributions without treating normal variation as failure.

05A practical example

Example

A checkout team defines start as entering development and finish as production release. Its 85th-percentile cycle time is 12 days, while the median is five. The longest items wait for security review, so the team adds earlier review slots and tracks whether the upper percentile improves over the next several releases.

06Where it breaks

Cycle time is a lagging measure for completed items and says nothing by itself about value, quality, or customer demand. Averages can hide wide variation. Changed workflow boundaries, item size, or data hygiene can create apparent improvement without faster delivery, and local optimization may merely move waiting elsewhere.

Common mistakes

  • Changing start or finish definitions while comparing periods as though the data were equivalent.
  • Using only an average and ignoring percentiles, outliers, and the shape of the distribution.
  • Pressuring individuals to reduce cycle time instead of investigating conditions in the system.

07Sources

  1. The Kanban Guide (May 2025)Kanban Guides · Official guideThe Kanban Guide (May 2025)
  2. Configure Lead Time and Cycle Time WidgetsMicrosoft Learn · Technical documentationConfigure Lead Time and Cycle Time Widgets
View saved topics

Start typing to search titles, definitions, categories, and frameworks.