Cycle Time

The gap between the two clocks is the real insight, not either number alone. A tiny cycle time next to a huge lead time says the problem is the queue, not the speed of the work. The fix differs. Asking people to work faster will not help when the real issue is capacity or priority, while clearing the backlog or routing work better will.

Cycle time also shows how much real capacity a process uses. That matters before taking on more customers or deciding what to automate. As a rough rule, the average amount of work in progress equals throughput multiplied by lead time, so a long queue and a slow flow tend to arrive together.

• Measure from when work starts to when it finishes, never from when the request arrived.
• Put it next to lead time for the same process to see how much of the wait is queue and how much is work.
• If cycle time is small and lead time large, look at capacity and priorities, not at how the work is done.
• Use it to estimate how many tasks a team can absorb before it needs more people.
• Watch for it creeping up while lead time looks stable. That can mean a task is quietly getting harder.
• Automate the steps with the longest cycle time first, where automation returns the most capacity.

Every request has two clocks. Lead time starts when the request arrives and ends when it is delivered, so it is what the customer feels. Cycle time starts only when someone begins the work and stops when it is done. The difference between them is waiting.

Take a support reply. A ticket arrives on Monday morning and is picked up on Wednesday. The reply takes ten minutes to write. Cycle time is ten minutes. Lead time is two days.

• Measuring from the date the request arrived, which gives lead time and hides where the delay sits.
• Averaging across very different tasks. A five-minute task and a three-day task need their own figures.
• Pushing people to shorten cycle time when the delay is in the queue.
• Cutting cycle time by skipping checks, which moves the cost into rework.
• Counting a task as started when it is assigned, not when work actually begins.
• Reporting the figure once. A trend over several weeks shows what a single number cannot.