Process time is how long one unit is being worked on at a step. Cycle time is how often a unit comes out of that step. Compare cycle time with takt, never process time: a machine with an 84-second process doubled up gives a 42-second cycle, which meets a 45-second takt.
The two words, as we use them
Stand at a step with a stopwatch. Process time is how long a unit is under work, and cycle time is the gap between one finished unit and the next.
With one machine doing one unit at a time and no waiting, the two match, which is why people swap the words. They split as soon as a step has parallel machines, batches or waits.
Definitions vary. Some books use process time for the whole route, or for value-adding work only, and some plants say cycle time when they mean process time, so write yours on the sheet.
Picture it with two machines
Two machines each take 84 seconds on a part, and they start 42 seconds apart. So a finished part comes out every 42 seconds, while each part still spends 84 seconds under work.
Where takt fits in
Takt is the pace demand asks for, and it has no process in it at all. Takt is set by the customer, cycle time by the step as it runs, and process time by the method and the machine.
Only cycle time can be set against takt, because both describe how often a unit must appear. To set takt, use the takt time calculator, and for lead time alongside these two see takt time vs cycle time vs lead time.
Worked example: a three-step cell
Example: a hypothetical cell has 7.5 available hours, or 27,000 seconds, and demand for 600 parts. Takt = 27,000 / 600 = 45 seconds.
| Step | Process | Resources | Planned cycle | Against 45 s takt |
|---|---|---|---|---|
| Wash | 600 s | 1 washer, 20 per load | 30 s | Meets it |
| CNC | 84 s | 2 machines | 42 s | Meets it |
| Test | 38 s | 1 station | 38 s | Meets it |
This is the classic trap. Someone sees 600 seconds on the washer against a 45-second takt, and asks for 13 more washers, when one load of 20 is enough.
Process time still does a useful job, because it tells you how much resource a step needs: process time / takt. That gives 84 / 45 = 1.87, so 2 CNC machines, and 600 / 45 = 13.3, so a load of at least 14 parts.
Planned cycle time vs measured cycle time
The planned cycle assumes a step always has work and space to pass it on. The measured one is run time / units made, and it includes every short wait.
Say the test station takes 38 seconds per part under work, and yet a shift average shows a measured cycle of 47 seconds. The extra 9 seconds per part are micro-stops, retests or waits.
At 47 seconds the cell can make 27,000 / 47 = 574 parts, which is 26 short of 600. Stopwatch studies of the process alone would never have shown this, so what are micro-stops is the place to look next.
In OEE terms, this gap between process and measured cycle is a performance loss, as long as the 38 seconds is the proven ideal. See speed losses and the six big losses calculator.
Which number for which job
- How many machines or stations do I need? Process time / takt.
- Can this step keep up with demand? Cycle time, measured, against takt.
- Which step limits the line? The longest measured cycle time, as in bottleneck analysis.
- How much work is in one unit? The sum of process times: 600 + 84 + 38 = 722 seconds per part.
- How long does a part stay in the cell? Throughput time, which adds the queues.
How to measure each one
- Process time: time the part under work with a stopwatch, from start of work to end of work, over 10 or more parts.
- Planned cycle time: calculate it from process time and resources, using the formula above.
- Measured cycle time: take run time / units made over a whole shift, using the cycle time calculator.
- Machine or person? If a person's time matters more, split it out, as in cycle time vs touch time.
Errors that buy the wrong machine
- Comparing process time with takt. It overstates the machines you need at parallel and batch steps.
- Dividing by machines that do not run together. A second machine that sits idle without an operator gives you no 42-second cycle.
- Ignoring load time on batch steps. If loading the washer takes 60 seconds, the washer's time per load is 660 seconds, so the cycle is 33 seconds.
- Assuming parallel machines are the same speed. If they differ, the cell rate is the sum of their rates, as OEE on parallel machines shows.
- Treating a stopwatch figure as the real cycle. Spread matters too, as cycle time variation explains.
The gap between process time and measured cycle time rarely shows on paper. Fabrico reads counts, speed and stops live from PLCs or from external sensors on older machines, and shows OEE per line and shift. See how Fabrico connects machines.
What people ask next
What is the difference between process time and cycle time?
Process time is how long a unit is being worked on at a step. Cycle time is how often a unit comes out of that step. They match only with one machine, one unit at a time and no waiting.
What is the difference between takt time, cycle time and process time?
Takt is the pace demand asks for, cycle time is the pace the step delivers, and process time is the work on one unit. Compare cycle time with takt, and use process time to size resources. See takt time vs cycle time vs lead time.
Can cycle time be shorter than process time?
Yes, with parallel machines or batches. Two machines at 84 seconds give a 42-second cycle, and a washer taking 600 seconds for 20 parts gives a 30-second cycle.
How do I calculate how many machines I need from takt?
Divide process time by takt and round up. An 84-second process against a 45-second takt gives 1.87, so 2 machines. Then allow for losses, as how to increase production capacity explains.
Is process time the same as ideal cycle time?
Often, for one machine doing one unit, though not always. Ideal cycle time in OEE is the proven best time per unit for the whole resource. On parallel machines it may be calculated per machine, so see OEE on parallel machines.
Fabrico is a light MES for OEE. It takes stops, speed and counts from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and operators tag the cause at the line.