Bottleneck analysis example: finding the slow step in a five-step order process

Capacity per step is 60 ÷ cycle time × workers. The step with the lowest capacity is the bottleneck: here Shipping at 33.3 orders an hour against 30 demanded. Adding a picker changes almost nothing; adding a shipper cuts the time an order spends in the process from 26.8 min to 19.3 min.

Bottleneck33.3/hShipping, 90% loaded at 30 orders an hour
One more picker26.1 mintime in process, from 26.8 min: inside the margin of error
One more shipper19.3 mintime in process; the bottleneck moves to Packing

The process

An order process has five steps in a row. 30 orders arrive every hour at random. The cycle times and workers below are illustrative numbers, not data from a real site; replace them with your own to repeat the analysis.

Step 1: capacity of each step

capacity per hour = 60 ÷ cycle time (min) × workers
Capacity and load of each of the five steps at 30 orders an hour
StepCycle timeWorkersCapacity per hourLoad at 30/h
Order receiving1.2 min150.060%
Picking4.5 min340.075%
Packing3.2 min237.580%
Quality check1.5 min140.075%
Shipping (bottleneck)1.8 min133.390%

Step 2: the lowest capacity is the bottleneck

Shipping can do 33.3 orders an hour, the lowest of the five. Nothing the other steps do can raise the output above that. At 30 orders an hour it runs 90% loaded, so it is not overloaded, but it has the least slack. Receiving, with 50.0 an hour, is the least loaded at 60%.

Step 3: confirm with a simulation

The capacity calculation ignores variation. Running the same five steps in the Simflowly engine (random arrivals, each cycle time lognormal with a coefficient of variation of 0.5, seed 7, 30 replications of 200 hours) gives the waits the arithmetic cannot:

Simulated utilization and average wait before each step, as is
StepBusyAverage wait before the stepAverage line
Order receiving60%1.1 min0.6
Picking75%1.8 min0.9
Packing80%2.6 min1.3
Quality check75%1.5 min0.8
Shipping90%7.6 min3.8

Shipping is the busiest step and the one with the longest wait in front of it: 7.6 min. The capacity calculation and the run name the same step. An order spends 26.8 min in the process on average (95% interval 25.9 min to 27.7 min), against 12.2 min of actual work.

Step 4: two fixes, one that works

Same demand, same seed, one extra worker in two different places:

Time in the process and wait at Shipping for three staffing choices
StaffingBottleneckTime in the process (95% interval)Wait before Shipping
As isShipping (33.3/h)26.8 min (25.9 min to 27.7 min)7.6 min
One more picker (not the bottleneck)Shipping (33.3/h)26.1 min (25.2 min to 27.1 min)7.8 min
One more shipper (the bottleneck)Packing (37.5/h)19.3 min (19.1 min to 19.6 min)3 sec

The extra picker reduces the wait in front of Picking, but orders still queue at Shipping, so the time in the process stays at 26.1 min: the interval overlaps the as-is interval, so the difference is not distinguishable from run-to-run noise. The extra shipper cuts the wait before Shipping to 3 sec and the time in the process to 19.3 min, because it raised the capacity of the step that was limiting everything. The bottleneck now sits at Packing, with 37.5 orders an hour, which is the next place to look.

What to take from this

  • Capacity first, then a run. The arithmetic names the bottleneck in a minute; the run shows how long things wait.
  • Do not add capacity at a step with slack. It shortens that step's own queue and barely changes what the customer experiences.
  • Expect the bottleneck to move. Each fix hands the limit to the next-slowest step.

Try it with your own numbers in the bottleneck calculator, or open the order picking template to run a similar process.

Questions

How do you find the bottleneck in a process?

Compute each step’s capacity as 60 divided by its cycle time in minutes, times its workers. The lowest capacity is the bottleneck, and it caps the output of the whole process. Then confirm with a run: the bottleneck is usually the step with the highest utilization and the longest wait in front of it.

Does the bottleneck always have the longest queue?

Not always. Here Shipping is the bottleneck at 90% busy and an average wait of 7.6 min, the longest of the five steps. Queues also depend on variation and on how workers are split, so use the capacity calculation to name the bottleneck and the run to confirm it.

If I fix the bottleneck, what happens next?

The bottleneck moves. After adding a second shipper, Shipping drops to 45% busy and Packing becomes the slowest step at 37.5 orders an hour. Capacity then rises only to that step’s limit, so find the next bottleneck before you plan the next fix.

Keep going