The sales lead who shows up as two boxes
A packaging printer measures the path from incoming order to order confirmation. The bottleneck is the credit check. The second finding is a role two steps share, and it only becomes an action once the first one is dealt with.
Approval and exception handling are two boxes in the flow and one person. Measured separately they read 57 % and 20 % utilisation, together 77 %, and above 100 % on a Monday. The finding is correct, and splitting that role would still be the wrong first move: as long as the credit check sits in front of it, splitting buys 0.05 working days.
0.6working days
In 8 of 10 cases between 0.2 and 1.7 days.
Bonitäts- & Kreditprüfung
In 78 of 100 simulated runs this step was the hold-up.
€48,000
Per year. Estimated between €43,000 and €54,000.
€37,000
Per year. Estimated between €32,000 and €42,000.
Order handling
- Industry
- Packaging printer
- Size
- 130 Mitarbeitende
- Cases per year
- 4,400
- Fully loaded rate
- €66 / h
- Simulated runs
- 120,000
- Date
- September 2026
Sales promises order confirmations the same day. Customers ring anyway to ask where theirs is. The sales desk says orders are released too late. Accounts receivable says it checks as fast as it can. The sales lead says she signs off every day. All three are right.
4,400 orders a year, six steps, three roles, a €66 fully loaded rate. Orders do not arrive evenly: on Mondays the weekend is sitting in the inbox, and at the start of the month the framework agreements call off. On the busiest day almost twice as many orders arrive as on an ordinary one. What is measured is the time from incoming order to dispatched confirmation, including the calendar waits: the credit agency's reply in unclear cases and the customer's response on an exception. Modelled and simulated over 120,000 runs.
Six steps, four roles, one measurable path
The process modelled in FlowVisual. ↯ marks a media break — the point where data is retyped from one system into the next.
| Step | Role | System | Duration P10–P90 | Bottleneck |
|---|---|---|---|---|
| 01Auftrag erfassen | Vertriebsinnendienst | E-Mail / ERP↯ | 6–18 min | 0 % |
| 02Bonitäts- & KreditprüfungBottleneck | Debitorenbuchhaltung | Auskunftei-Portal / ERP↯ | 6–36 min | 78 % |
| 03Preis & Auftragsbestätigung | Vertriebsinnendienst | ERP / Excel↯ | 10–28 min | 5 % |
| 04Freigabe Vertriebsleitung | Vertriebsleitung | ERP | 2–8 min | 10 % |
| 05Klärfall entscheiden | Vertriebsleitung | E-Mail / ERP↯ | 5–16 min | 7 % |
| 06Auftragsbestätigung senden | Vertriebsinnendienst | ERP / E-Mail | 4–11 min | 0 % |
120,000 runs, one clear answer
Every chart below shows before against after. Values are labelled directly — the colour is a second signal, never the only one.
100
Aufträge/Woche
3 %
0.2–1.7working days
Bonitäts- & Kreditprüfung — 78 %
Start with the spread, because the promise hangs on it. Median lead time is 0.6 working days, the P90 is 1.7. Turn that median into “confirmation the same day” and you are wrong on one order in ten, by more than a full day. In fact 10.6 % of confirmations go out later than promised.
In 78 % of runs the bottleneck is the credit check. It is one person, and every order passes through her: 88 % utilisation on average, 118 % on a Monday, 171 % on the first Monday of the month. From roughly 85 % onwards the queue grows faster than the load, and that is exactly where she sits. The “days over capacity” figure stays low at 3 %. That is not an all-clear, it is a question of resolution: it only counts backlogs larger than a full day's work. This backlog builds on Monday and is cleared by Thursday. It is felt regardless, just inside the week.
The second finding has no row of its own, and that is its whole problem. Approval (step 04) and exception handling (step 05) are the same person. In the bottleneck table they sit far down the list at 10 % and 7 %. Together they bind 17 % of runs, making them the second constraint in the model, without any single figure anywhere reading 17 %.
The arithmetic behind it is simple and still the point. The sales lead gives this process 170 minutes a day. Approval demands 98 of them, exception handling 33, together 131. Score each step separately against that same day and you read 57 % and 20 %: two numbers nobody questions, because both sit well below the level at which a queue forms. It is the same work, only attributed differently. Together it is 77 %, on a Monday 103 %, and on the first Monday of the month 150 %. And if that one person is out for a day, both steps stop, not one.
Then the loop: 16 % of approvals go back to pricing as an exception rather than onward. Pricing and approval therefore run 1.19 times per order, the exception step 0.19 times, which is 835 times a year. What all of it costs is roughly 583 hours of avoidable work a year, mostly inside the check itself and in answering status calls, plus around €10,100 of express surcharges for orders whose production slot can only be held at a premium after a late confirmation.
The credit report comes through an interface.
The obvious intervention would have been a different one. Once the shared role has been named, you want to split it: approvals to a second manager, exceptions stay with the first. The finding is right. At this point the action would not be.
Run through the model, splitting the role on today's situation would buy 0.05 working days at the median, a little over twenty minutes, and a saving whose range runs from minus €500 to €3,000, which makes it indistinguishable from zero. Meanwhile the credit check's bottleneck probability would rise from 78 % to 85 %: the second constraint had been taking runs off it, and without that shadow the first one stands there unchanged and louder.
So the intervention goes in front of it. The credit agency is queried automatically from inside the ERP. A person only decides the cases where no clean match comes back or the limit is exceeded, which is 22 % of orders. For the rest the check falls from 6 to 20 minutes down to 2 to 5. The wait on the agency stays in place for the unclear cases; it does not belong to the company.
Everything else would stay as it is: same roles, same loop, same approval by the same person. The intervention would not be free. It is costed at €12,000 of setup plus €2,600 a year for the interface and query fees; the running cost is already deducted from the saving shown. Payback would sit at around four months.
Median lead time down 50 %
- How long it takes
- 0.6 → 0.3 working days
- Throughput
- 100 → 114
- Days over capacity
- 3 % → 0 %
- Saving per year
- €37,000
Median lead time would fall from 0.6 to 0.3 working days, a drop of 41 %, and the P90 from 1.7 to 1.3. The share of confirmations going out later than promised would fall from 10.6 % to 6.1 %, express surcharges from around €10,100 to €6,000 a year. The credit check would fall from 88 % to 45 % utilisation, and on a Monday from 118 % to 61 %.
The bottleneck moves, and it moves onto the shared role. Approval alone would reach 43 % bottleneck probability, together with exception handling 51 %. Before, it was 17 %. The bottleneck would no longer be a station, it would be a person, and in the table she would still appear as two rows.
And now the split becomes an action. Applied to the state after the intervention, median lead time would fall again from 0.33 to 0.23 working days, a drop of 30 %, and the P90 from 1.26 to 1.06. Applied to today's situation it was 0.05 days and 8 %. Same finding, same action, same arithmetic. All that changed is what sits in front of it.
That is the return on this analysis, and it appears in neither table: the right finding at the wrong time is not an action. A bottleneck list says where things stick. It does not say in which order to unstick them, and with two constraints in series that order is what decides the effect. Tackle the second one first and you get a true statement, an empty budget, and a first bottleneck that binds harder afterwards than it did before.
What the intervention deliberately does not touch: the exception loop. 16 % of approvals still go back to pricing, 835 times a year. It is the next candidate after the shared role, but only after a fresh measurement and with the same question: is anything still sitting in front of it?
What this analysis cannot tell you
Every measurement has limits. A measurement that hides them is advertising.
- 01
This is a sample analysis. Company, process, roles and timings are constructed from typical mid-market structures, not collected at a client. It shows what an analysis looks like. It does not show what your result will be.
- 02
The figures differ from the FlowVisual template for the same process, and that is not an error. First, these are two different example companies: the template computes 6,000 orders a year with its own capacities, this model 4,400 with different ones. Second, this analysis includes calendar waits the tool does not compute, namely the credit agency's reply in unclear cases and the customer's response on an exception. Both calculations are correct and answer different questions: one says how long the organisation takes, the other how long the work takes. What is comparable are the bottleneck shares, not the days.
- 03
“Days over capacity” here means: on that share of working days, at least one role ends the day with more work outstanding than it can clear in a day. It does not mean anyone worked overtime. That this figure is low at 3 % while one role runs at 88 % is normal for a process with no batching and no fixed signing slots: the backlog builds on Monday and is gone by Thursday, but it rarely exceeds a full day's worth of work.
- 04
The simulation serves every role in order of arrival. Priority handling is not modelled. A sales lead who always signs approvals first would have a better P90 and the same finding, because the 131 minutes do not change.
- 05
Shortening a credit check is a risk decision, not a lead-time decision. This analysis calculates the time gained and quantifies the share an interface would take over. It does not price what a wrong automatic release costs. The 22 % of unclear cases deliberately stay with a person; move that line further and you trade time gained against default risk.
- 06
Express surcharges are modelled at 2.5 % of order value, capped at €150, applying to 30 % of orders confirmed late. All three are assumptions; the costing carries them with a range of 0.6 to 1.4.
- 07
Setup and running costs for the interface are assumed vendor figures. A cheaper offer shortens the payback, a deeper ERP integration lengthens it.
- 08
“Throughput” is the ceiling the scarcest role allows, not the actual volume. That is 88 orders a week, so before the change, 88 % of the ceiling.
- 09
P10–P90 is not a worst case. In 10 % of cases it takes longer than the P90 value.
Your process will look different.
This analysis is a sample. Your numbers are not. With FlowVisual you model your own process and get the same evaluation — on your machine, with your values.