Bottleneck analysis: where does the process get stuck?
The bottleneck of a process is the role or department with the highest utilization, provided it reaches at least 85%, because it limits how many cases the process can handle. The bottleneck step is the step that puts the most load on that role. It is often not the step where cases wait longest; long waits can have other causes, such as a manager who looks at approvals once a week.
How the app determines the bottleneck
For every lane with a headcount, the app compares demand with capacity (see Staffing calculation). The lane with the highest utilization is the bottleneck if its utilization is at least 85%. Below that, the app names this lane only as the highest utilization, not as a bottleneck. Within that lane, the step with the largest load, visits times processing time, is the bottleneck step.
- Amber: utilization from 85% to 100%. Little slack, waiting times grow.
- Red: above 100%, or the bottleneck itself. Above 100% the app shows the backlog: the cases per period that are left over, counting only the cases that pass through this lane. For the whole process what counts is how many fewer cases per period reach an end than arrive – never a sum across lanes. The lead time of the instant calculation is then a minimum and marked “only without backlog”, because the actual lead time grows every period.
- Only roles and departments are considered. External partners, customers and IT systems count towards the lead time, but they are not reported as bottlenecks, because their capacity is not in your hands or not measured in people.
Worked example: the longest wait is not the bottleneck
In the invoice example, the approval by division management waits 1,440 minutes, three working days, before anyone looks at it. In a workshop that is the step everybody complains about. Yet division management is at 84%: at full inflow it would need 25 hours of approvals a month against 29.3 hours available, but because accounts payable leaves 10 invoices a month, only 297 approvals arrive, 24.8 hours. Its long wait does not come from overload. Accounts payable has short waits, but it is at 101%: 266.7 hours of work against 264 hours of capacity. Accounts payable is the bottleneck, and the bottleneck step is "Enter invoice", the step with the largest load in that lane at 6 minutes per case.
| Lane | Longest wait per pass | Utilization | Status |
|---|---|---|---|
| Accounts payable | 120 min (posting) | 101% | bottleneck, red |
| Division management | 1,440 min (approval) | 84% | below the threshold |
| Supplier | 2,400 min (query) | – | external, not assessed |
What helps at a bottleneck
There are three kinds of lever, and the cheapest one usually comes first:
- Reduce the load per case: avoid queries, remove double entry, automate a step. In the example, making the order number mandatory halves incomplete invoices from 20% to 10%; accounts payable drops from 101% to 97.5% and the backlog disappears.
- Move work: let another role take over a step that does not need the bottleneck’s skills.
- Add capacity: more people or more availability for this process. This works, but it is usually the most expensive lever and the hardest to reverse.
A bottleneck moves
Once the bottleneck is relieved, another lane becomes the one with the highest utilization. In the example, after the measure accounts payable is still the highest at 97.5%, and division management rises from 84% to 85%, because now every approval arrives. Calculate again after each change instead of assuming the problem is solved. The app recalculates with every change, and measures can be switched on and off one at a time to see their effect; more in Future process and measures.
The honest limit of the instant calculation
The instant calculation uses the waiting times you entered. It shows that a lane is overloaded, but it does not make the waiting time in front of that lane grow by itself. In reality an overloaded team creates queues, and the lead time rises. The simulation in the analysis area does exactly that: it sends cases through the process with limited staff, spreads processing times between best and worst case and shows queue lengths and backlog over time. Use the instant figures to find the bottleneck and the simulation to see how severe it is. The differences are explained in Simulation and process mining.
Questions for the mapping session
To find a bottleneck you need three figures per role. Ask:
- How many people do this work, and how much of their time goes into this process?
- How long does the step take when you work on it without interruption?
- How many cases come in per day or per month?
Try it in the editor
Invoice check (worked example) · 1,000 / month
In the worked example accounts payable is marked red at 101% and division management is at 84%; switch on the measure to see the bottleneck relieved.
Open the example in the editorFrequently asked questions
Why is no bottleneck shown?
Either no role or department has a headcount, or the highest utilization is below 85%. In the second case the app names the lane with the highest utilization instead. Enter people and availability for the roles involved.
Can an IT system be a bottleneck?
Not in the instant calculation. IT systems, customers and external partners are not given a utilization. If a system is slow, enter the waiting time it causes.
Is the bottleneck the same as the step with the longest lead time?
No. Long waits often come from organization, such as approval rounds once a week. The bottleneck is about capacity: where demand comes closest to or exceeds available time.
Self-test
Keep learning
Would you like to map or improve a process together with us? Contact