Stage 2 · Analyse · 04 · 25 min
Measure where the time and money go
Opinions about a process end where the numbers start.
This module measures the custom order process by hand: how long an order really takes, how much of that is work, where capacity runs out, and what the customer actually experiences. It ends with the baseline every later change is measured against.
Apply
1. Calculate time through the model
What you do here: calculate cycle time through a gateway, a parallel split and a rework loop.
Four rules cover almost every BPMN model:
| Pattern | Rule |
|---|---|
| Sequence | Add the times |
| Exclusive gateway (XOR) | Weight each path by its probability |
| Parallel gateway (AND) | Take the longest path; the others run alongside |
| Rework | Add probability × the extra time. If a step repeats until it passes, its expected time is T ÷ (1 − r) |
Apply them to Northstar’s custom order process.
Figures from the as-is model and last quarter's orders.
| Element | Data |
|---|---|
| Engineering review, small variants (60 % of orders) | 0.2 days of work |
| Engineering review, redesigns (40 %) | 3.4 days of work |
| Typical order without rework | 28 days, quote to invoice |
| Failed final inspection | 15 % of orders; each failure adds 18 days |
| Engineering review incl. its wait, then order materials | 5.5 days, then 1.5 days, in sequence |
What is the expected engineering review work per order, in days?
Apply
2. How much of the time is work?
What you do here: compute cycle time efficiency and decide which lever moves lead time most.
Cycle time efficiency is work time ÷ lead time. In office and engineering processes it is often far below 50 %, which means the lever is the waiting, not the work.
The typical custom order from the as-is model.
| Measure | Days |
|---|---|
| Lead time, quote to invoice | 28 |
| Work on all steps | 10.5 |
| Waiting between steps | 17.5 |
| Manufacture (the largest block of work) | 5 |
| Wait before engineering review (the largest wait) | 4 |
What is the cycle time efficiency, in percent?
Apply
3. Find the bottleneck
What you do here: find the resource with the least spare capacity and explain the queue in front of it.
Utilisation is demand ÷ capacity. The resource with the highest utilisation is the bottleneck. As utilisation approaches 100 %, waiting grows steeply, not gradually: there is no slack left to absorb a busy day.
Hours of work demanded by custom orders and other work, against hours available.
| Department | Hours demanded | Hours available |
|---|---|---|
| Sales | 540 | 960 |
| Engineering | 1380 | 1440 |
| Planning | 270 | 480 |
| Production | 4800 | 6000 |
| Quality | 360 | 480 |
What is Engineering's utilisation, in percent?
Apply
4. Look past the average
What you do here: read the spread of lead times and decide what to report.
An average hides the orders customers remember. Report the median and the 90th percentile, and split by order type.
Quote to invoice, in days.
| Order type | Share | Median | 90th percentile |
|---|---|---|---|
| Small variant | 60 % | 27 | 31 |
| Redesign | 40 % | 30 | 35 |
| Any order with rework | 15 % | 46 | 52 |
| All custom orders | 100 % | 28 | 44 |
Average across all custom orders: 30.7 days.
Small variants need 0.2 days of engineering, redesigns 3.4. Why is their lead time almost the same?
One more thing the spread tells you. Work that is uniform, with clear rules and little variation, can follow a fixed path. Work that varies a lot needs judgement. Small variants are the first kind, redesigns the second. You will use this in module 6.
Apply
5. Fix the baseline
What you do here: calculate the baseline for time, reliability, quality, cost and satisfaction, from last quarter’s counts.
Last quarter: 180 custom orders.
| Dimension | Raw data | Baseline |
|---|---|---|
| Time | Median lead time, quote to invoice | 28 days (90th percentile 44) |
| Reliability | 104 of 180 orders shipped by the promised date | ? |
| Quality | 153 of 180 orders passed final inspection first time | ? |
| Cost | EUR 223,200 process cost (excluding materials) for 180 orders | ? |
| Satisfaction | Survey after delivery, 1 to 5, 64 responses | 3.4 |
On-time delivery, in percent?
Freeze these numbers and their definitions now. If the definition changes later, the comparison is worthless.
Reflect
Check your understanding
Three questions on the module.
Two branches after a parallel gateway take 3 and 5 days. How long does the parallel block take?
Take away
What to keep
- Sequence adds, XOR weights, AND takes the longest, rework adds probability × extra time.
- Cycle time efficiency shows whether to speed up work or remove waiting.
- The bottleneck is the highest utilisation. Near 100 %, queues explode.
- Report median and 90th percentile, by order type.
- Freeze the baseline and its definitions before anything changes.
After this module you can
- Calculate cycle time through sequences, gateways and rework loops
- Compute cycle time efficiency and pick the right lever
- Find the bottleneck with utilisation and Little's law
- Report lead time beyond the average
- Set a baseline for time, reliability, quality, cost and satisfaction