How to find business process bottlenecks

01

Follow work from request to completion

Choose a process with a clear start and finish, such as preparing a quote or fulfilling an order. Walk through a sample of recent items with the people doing the work. Record each step, owner, handoff, queue and approval, including rework and exceptions.

Use timestamps from existing systems where they are reliable. If they are missing, observe a sample and label the measurement period. Note changes in demand, staffing and work complexity that could affect the comparison.

• Count items arriving and completed during the same period.
• Measure active work time separately from time waiting in a queue.
• Record work returned for correction and the reason it returned.

02

Compare demand with capacity

Throughput is the number of completed items in a period. A growing queue at a step, combined with sustained high use of its available capacity, is a signal to investigate that step. Check downstream blocks and quality problems before calling it the constraint.

Illustrative example: a team can complete 60 standard reviews per week, while 75 arrive each week. If the mix and capacity stay the same, unfinished work grows by about 15 reviews each week. Making the preceding intake step faster would not resolve the review queue.

An average can hide urgent work and unusually complex cases. Compare like with like and watch the process over more than one cycle; the constraint can move as demand or staffing changes.

03

Change one constraint and observe

Ask the people doing the work why items wait. Remove an unnecessary approval, clarify the information needed at intake or rebalance work only after checking the effect on quality and accountability.

Agree a baseline and a review date. Track completed items, queue age, turnaround time and rework together. If throughput rises but errors or overdue work rise too, investigate before expanding the change.

• Document the current process and who can approve a change.
• Try the change on a defined scope and keep an exception path for unusual work.
• Review whether a different step has become the limiting constraint.

04

Check before adding people or software

Observe the suspected step over several representative periods. Record arrivals, completions, oldest item waiting, available staff time and reasons the team could not proceed. Separate work waiting for information from work waiting for an available person; the remedies may differ.

Check whether the step is actually limiting finished work. A queue may build because an earlier team releases work in batches or because a downstream team cannot accept it. If the suspected step has spare time while customers still wait, look for missing inputs, uneven work mix or a different constraint before buying capacity.

After a trial, compare completed, acceptable work over the same type of period. Review queue age alongside quality and overtime. If the backlog moves to another step, update the process map and reassess rather than continuing to optimize the original step.

• Observation: arrivals, completions, queue age, work mix and unavailable time.
• Decision: the step most likely to limit accepted completions and the evidence for that judgment.
• Follow-up: the next constraint, rework and any service or quality trade-off.

05

A real example of finding the constraint

In a case published by NIST, the manufacturer Stainless Works worked with the nonprofit MAGNET to review its scheduling process. The team found that the production routes in its planning system contained incorrect information. After updating them, it identified welding as the main bottleneck and scheduled that constrained department differently.

NIST reports that the company reduced lead times from 12–14 weeks to five weeks and recorded increased or retained sales. This was a manufacturing project by Stainless Works and MAGNET, not a Capacity Expansion client engagement; the result does not predict what another business will achieve. The transferable lesson is to verify the information behind the schedule before treating the schedule itself as the problem.