Quality Management: When Inspection Becomes the Bottleneck

It is 14:00. Four machining cells finish their batches at almost the same moment, right on schedule. Every batch needs a dimensional check on the coordinate measuring machine (CMM) before the next operation can start. There is one CMM, and each check takes 45 minutes. The last batch will not be released until 17:00, after the second shift has already set up for something else.

Quality management in manufacturing is the system a plant uses to make sure products meet requirements: planning the checks, doing the inspections, deciding what happens to parts that fail, and fixing the causes. This article is about the part planners feel every day: inspection is a resource with capacity, and when it is full, your fastest machine waits.

Inspection is a machine too

Think of airport security. Four flights land at once, there is one scanner, and it does not matter how fast the planes flew. Passengers leave the airport at the speed of the checkpoint.

The CMM works the same way. Here is the queue at 14:00:

• Batch 1: waits 0 min, done at 14:45
• Batch 2: waits 45 min, done at 15:30
• Batch 3: waits 90 min, done at 16:15
• Batch 4: waits 135 min, done at 17:00

Average wait: (0 + 45 + 90 + 135) ÷ 4 = 67.5 minutes, for a check that takes 45. The cells finished on time. The parts did not move on time.

If the schedule had staggered the finishes by 45 minutes, the average wait would be zero, with exactly the same machines and the same inspector.

Why a "busy" inspection station waits so much

The 14:00 pile-up is a planned collision. In real life, batches arrive at random, and a second effect kicks in: waiting time explodes as the station approaches full load.

A simple textbook queue model (one server, random arrivals) shows the shape. Average waiting time is roughly utilization ÷ (1 − utilization) × check time. With a 45-minute check:

• At 50% busy: 1 × 45 = 45 minutes of waiting
• At 80% busy: 4 × 45 = 180 minutes
• At 90% busy: 9 × 45 = 405 minutes, almost 7 hours

The curve is real even where the formula isn't exact. Going from 80% to 90% busy does not add 10% more waiting. It more than doubles it. That is why an inspection station that "has spare capacity on average" can still be the reason orders ship late.

First-pass yield: the hidden 10%

The second number every planner should know is first-pass yield (FPY): the share of parts that pass every step the first time, without rework or scrap.

Say a part goes through three operations with these pass rates: 95%, 97% and 98%. Each looks fine alone. Together:

0.95 × 0.97 × 0.98 = 0.903, or about 90%.

Out of 1,000 parts started, around 903 go straight through. Around 97 need rework or become scrap. Nobody sees a "10% problem" on any single report, because each step reports its own good-looking number.

How rework hides demand from the schedule

Those 97 parts don't disappear. They go back to a machine, often the bottleneck, and use time the schedule never planned. If each rework takes 6 minutes on the grinder, that is 97 × 6 = 582 minutes, almost 10 hours. More than a full shift of capacity nobody asked for.

The schedule still shows the grinder as 85% loaded. Real load is higher, so the rest of the week slips, and people blame the schedule. The better answer: record rework as real operations with real times. Then the plan knows about them, and you can see what they cost. Scheduling that respects a single shared resource, like the CMM or the grinder, is covered in production scheduling and, for many routes crossing the same machines, in job shop scheduling.

FPY also connects to equipment numbers. The quality factor in OEE is the same idea for one machine: good parts divided by total parts.

Myth: "Quality slows production down"

The belief: inspection is the brake, production the engine. Unplanned quality slows production down. Planned quality speeds it up.

• A check scheduled as an operation, on a resource with a calendar, becomes a known 45 minutes.
• A check nobody planned becomes a surprise queue.
• A defect caught at incoming inspection costs one pallet. The same defect caught after three machining steps has used three machines' time. That is why incoming inspection of raw materials pays back so fast.

Four practical fixes

1. Put inspection in the routing as an operation with a time and a resource.
2. Stagger arrivals at shared inspection, 30–45 minutes apart.
3. Keep inspection below full load, or move simple checks to the operator.
4. Track FPY per product, not per step.

When a batch fails, the question "which other batches used the same material?" follows right away. That answer comes from batch traceability.

How it looks in factory.online

In factory.online, an inspection step can be an operation in the routing on its own resource, so the CMM appears on the finite-capacity Gantt next to the four cells and the 14:00 collision is visible before it happens. Quality inspections can put a batch on hold, output and scrap confirmations record what passed and what didn't, and rework can be planned as work on the schedule.

BEEDIGIT links quality checks, accepted and rejected output, and traceability records to shop-floor work. This adds production context when investigating defects or deciding which quantities can move to the next operation.

If your fastest machine spends its afternoons waiting for a measurement, book a demo and bring that week's schedule along.