I approve powder processing equipment for a living. And sometimes I don't.

I'm a quality manager at Hosokawa, which means anything we ship — mixers, mills, classifiers, granulators — goes through my release process first. That's roughly 200 unique systems and major upgrades a year. In 2024 I held back about 6% of first builds for rework. 5.8%, to be exact.

Every one of those rejections cost us something: parts, labor, schedule, or all three. That's fine. Every one of them would have cost a lot more if I'd let it ship.

Here's the opinion that gets me in trouble with our production team: Prevention isn't a delay. It's a shortcut.

The most expensive sentence in powder processing

'We'll fix it on site.' I hear that sentence a few times a year, usually when a schedule is slipping. Sometimes it comes from a customer. Sometimes from a sub-supplier. And once in a while, if I'm honest, from one of our own engineers who wants to hit a delivery date.

The sentence sounds practical, but it ignores what changes once the equipment is installed. On our floor, a leaking shaft seal during a test run means I stop the test, a technician grabs the right tool, and the right part is twenty meters away in stores. An hour later, we're running again. On your floor, the same leak means shutting down a line. It means coordinating with the production schedule, waiting for a service visit, and deciding whether to air-freight a part that weighs more than you want to think about. A two-hour fix becomes a two-day event.

In 2022, a customer asked to skip the factory acceptance test on a mixing system so delivery could move up four days. We recommended against it. They signed a deviation note and the machine shipped. It ran fine for nine days, and then a gearbox bearing failed. We couldn't have seen the defect in a static check — but the extended test we recommend, the one they opted out of, cycles the drivetrain under load for hours and would very likely have caught it. That test cost $2,400. The emergency service call, replacement bearing, freight, and downtime came to about $18,000. That customer now puts 'FAT required' in every equipment contract they sign.

Let me be fair: the skipped test didn't break the bearing. What it did was remove the last cheap chance to find it.

What a factory acceptance test can't tell you

I also don't want to oversell tests. What most people don't realize is that a factory acceptance test is designed to prove a machine runs — not that your process will run. It can't. Your material, your feed consistency, your upstream pressure, even the humidity in your building affect how powder behaves. When we test a Nauta mixer or an Alpine classifier in our shop, we're testing the machine hard, with margins built in. We're not pretending to be your production line.

Here's a truth that doesn't make it into many sales brochures: the thing that most often delays a commissioning isn't a broken part. It's documents that don't match the machine. The P&ID shows one valve arrangement; the installed piping has another. The motor nameplate doesn't match the amp draw in the test report. Your team will find those discrepancies sooner or later — but 'later' usually happens at 11 pm on a startup day. That's why I treat the documentation package as a deliverable, not an afterthought.

Three checkpoints worth paying for

So what actually prevents commissioning problems? In my experience, it's not one heroic final inspection. It's a series of deliberate, somewhat boring checkpoints. If I could make every customer demand three things from their equipment supplier, it would be these:

1. A design review before the steel is cut. Not a brochure presentation — a review of material of construction, seal materials, surface finish, and clearances against the actual application. This is the cheapest place to change a decision, and the place where most people don't spend enough time.
2. Verification at sub-assembly, not just at the end. Motor alignment, weld quality, component certifications — those checks are dramatically easier before the machine is fully assembled. Once a gearbox and motor are bolted in place and guarded, you're not checking alignment; you're checking your eyesight.
3. A factory acceptance test that includes time under load. A machine that runs for 20 minutes can hide a lot. A machine that runs for four to eight hours under load, with logged vibration, temperature, and current readings, is harder to fool. Pair that with an as-built documentation review, and you have the best picture a vendor can give you before you commit to shipping.

Notice I didn't list 'final inspection' as a separate item. Final inspection matters, but it's the last gate — not the whole fence.

The 'we'll tune it on site' idea is mostly nostalgia

There's a romantic image of the engineer who arrives on site, listens to a machine for 20 minutes, and adjusts it by feel. I respect that image because I've known people who could do it. But it comes from an era when powder processing equipment was simpler — when you could loosen a locking nut, move a lever, and watch the result on a gauge right in front of you.

That was true 20 years ago. Most modern systems are different. Parameters are set through a control system, logged, and protected. Maybe a small adjustment on a classifier wheel or an air gap still happens by hand. But a 'quick feel' of a machine doesn't tell you much when the behavior is buried in a PLC program and a dozen interlocks. Tuning an unfamiliar machine by trial and error on a customer's line is exactly the kind of risk that a few hours in a factory test would have removed.

Yes, checking takes time. So does rework.

The most common pushback I hear — from customers and from our own sales team — is that inspection takes time. I won't argue with that. Adding checkpoints to the front of a schedule is about as popular as a diet plan in December.

But here's what I've watched happen over and over: a project is on schedule until it isn't. And when it slips, it's usually because a problem was found late. The schedule where issues are discovered early tends to be the one that ships close to on time. I've never seen an emergency service visit improve a delivery date.

In quality management there's an old rule of thumb called the rule of ten. The later you find a problem, the more it costs — roughly ten times more for each stage you let it pass. I don't know if the multiplier is exactly ten. What I do know is that a wrong seal discovered in week four costs a phone call, and the same wrong seal discovered after crating costs a crate, a crane, and a redo. The multiplier doesn't matter when the decision is already in front of you.

The other objection I hear is, 'we have our own commissioning team; they'll catch issues.' I don't doubt it. But a good commissioning team is not a reason to give them avoidable surprises. A site acceptance test is the most expensive place to discover a problem. The best outcome of any site acceptance test is that nothing interesting happens.

At the end of the day, my job isn't to make our release process look thorough. It's to make sure the machine doesn't come back to us as someone's expensive lesson. I've been signing off on powder processing systems long enough to know that every fix was cheaper to prevent before shipment than to correct after it was installed.

Prevention isn't a delay. It's the shortcut.