
The short answer
Look for the step that limits finished work.
A bottleneck is the part of a workflow that limits how much useful work the whole system can finish. The busiest person is not automatically the constraint. Look for a queue that keeps returning, then investigate what makes work wait there.
- Start with an outcome: shipped orders, qualified sales or decisions made—not how busy the team feels.
- Follow the waiting: trace the handoffs before assuming the answer is more people or more demand.
- Test one explanation: a useful change improves finished work without shifting the cost into rework or quality failures.
Monday’s dispatch list is still open on Thursday. Sales wants another campaign. Operations wants another person. The founder is approving exceptions between calls. Each request sounds reasonable. None establishes what is actually holding orders back.
Before choosing a solution, ask a narrower question: where does an otherwise ready order stop moving? That question gives the team something it can observe together.
A symptom tells you where to look. It does not tell you what to buy.
Late delivery is an outcome. A growing inspection queue is a clue. Missing inspection criteria, repeated defects or limited inspection capacity are different possible explanations. They require different responses.
| What you see | What to check | Evidence to collect | Avoid jumping to |
|---|---|---|---|
| Orders leave late | Where do ready orders wait longest? | Time at each handoff; completed dispatches per day. | Buying leads before delivery can absorb more work. |
| The pipeline is full, but sales are weak | Which stage loses qualified buyers, and why? | Stage conversion, time in stage and recorded loss reasons. | Calling every unclosed lead a demand problem. |
| Revenue grows, but cash gets tighter | Is cash tied up in stock, unpaid invoices or weak margins? | Collection dates, inventory age and contribution by order. | Assuming another sale will solve the cash gap. |
| Every decision comes back to the founder | Is authority unclear, or is information missing? | A log of recurring decisions, owner and waiting time. | Hiring a manager without defining decision rights. |
A queue alone is not proof of a capacity shortage. Work may arrive in batches, requirements may be incomplete, or a downstream team may release work only once a day. Compare several jobs and distinguish time spent working from time spent waiting.
Follow one delayed order, not every dashboard.
Hypothetical example · not a client case study
Imagine a small manufacturer that receives 30 similar orders each week. Production can prepare 30, but final inspection releases only 20. Ready orders accumulate before inspection while dispatch waits for clearance.
- 01Order confirmedRequirements agreed
- 02ProductionWork completed
- 03Final inspectionReady work queues here
- 04DispatchWaiting for release
Adding production capacity would make the queue larger if inspection stayed unchanged. But hiring an inspector is still only one possible answer. The team first needs to learn why inspection releases fewer orders.
Ask three questions at the queue
- Is the work ready? Check whether specifications, drawings and approvals arrive with the job. Missing information can consume the time that appears to be a staffing problem.
- Is the work right the first time? Separate first inspections from repeat inspections. If defects create most of the workload, adding inspection capacity may leave the cause untouched.
- Can the owner make the decision? Record which jobs wait for founder approval despite meeting a known standard. That is a decision-rights question, not necessarily an inspection-capacity question.
Test the explanation before funding the solution.
Suppose the manufacturer finds that inspection repeatedly pauses to clarify requirements. A small test could require a complete job pack before work enters production. Assign an owner to check readiness and record the reasons a job is held.
Choose the measure before starting: finished orders released, time spent waiting and first-pass quality. A shorter queue is not an improvement if the team has simply stopped accepting work or has skipped a required check.
- If completed work improves and quality holds: repeat the test over more jobs before standardising the change.
- If the queue stays: review the explanation. Capacity, scheduling or rework may still be limiting release.
- If the queue moves downstream: follow the new constraint. Improving one step does not guarantee the whole workflow is balanced.
Three mistakes that hide the real problem
Measuring utilisation instead of output. Keeping every person busy can create excess work in progress. Ask whether the customer-facing outcome improved.
Changing several things at once. A new hire, new software and a new meeting rhythm make it hard to know what helped. Start with a test small enough to explain.
Turning the investigation into blame. Track handoffs, requirements and decision rules. A person carrying the queue may be absorbing a problem created elsewhere.
Put it to work
Your seven-day constraint worksheet
Use one outcome and one workflow. These notes stay in this page, are not sent or saved, and disappear when you reload. Copy them before leaving.
Observe the work
Which outcome is slipping? Trace a few recent jobs from start to finish. Record where each waited and why.
Write a testable explanation
Name the step, the suspected cause and the evidence that could prove you wrong. Separate facts from assumptions.
Change one thing
Choose a small, reversible test with a named owner. Keep quality and safety requirements intact.
Review the whole flow
Compare completed work, waiting time and quality with the starting point. Decide whether to continue, revise or investigate elsewhere.
Choose the next step from what you found.
If work waits for instructions or approvals, examine founder dependency. If the work has no consistent method, start with a usable SOP. If the main question is still where to look, return to the Diagnose pillar.
The goal of this week is a better explanation and a small amount of useful evidence. A seven-day test is a starting point; seasonality, complex jobs and long sales cycles may need a longer observation window.


