Note

The Hidden Cost With No Line Item: What's Really Slowing Down Malaysian Factories

A sales manager promises a long-standing customer an earlier delivery date. Her screen shows an open slot on the schedule, so she says yes and hangs up, pleased to have kept a good client happy. What she doesn't know is that an hour earlier, the shop-floor planner two buildings over gave that same slot to a different job, after a phone call of his own, and hasn't had a moment to update the system yet. Nobody made a mistake. Two people just looked at two different, equally confident versions of the truth. By Thursday, someone has to find out, apologise to the customer, and rebuild two days of schedule by hand.

That lost time and that awkward phone call are a real cost. It just doesn't have a name on most companies' books.

What it actually is

Call it the orchestration tax: not the cost of gathering information — modern software does that well now — but the cost of actually acting on it once two parts of the business have quietly drifted onto different pages. Someone has to notice the gap, work out what to do about it, and do the coordinating. That someone draws a real salary, every single day, and almost no accounting system gives that work its own line item.

Instead, it hides inside other costs that look completely unrelated:

• An expediting fee paid to rescue an order that's about to be late
• Safety stock sitting in a warehouse purely as insurance against information that might be wrong or late
• A production run redone because someone built to a spec that had already quietly changed
• A good customer lost to a competitor who simply answered the phone faster

Each one has a real, traceable cause upstream — and it's almost always the same cause: two parts of the same business, operating for slightly too long on two different versions of reality.

Why it gets worse exactly as a company grows

In a six-person shop, this problem barely exists, because sales, production, and the books are all the same handful of people in one room — nobody needs telling something everyone can already see. Growth breaks that arrangement for an entirely unsentimental reason: no one person can hold sixty people's worth of jobs, machines, and customers in their head the way one person could hold six. So the work gets divided — a sales team, a planning function, a finance department, each eventually running its own software — and that division recreates, deliberately, the very separation that never used to exist. Every boundary it creates then has to be crossed by something: a meeting, a status report, a phone call, a system that's supposed to talk to another system and often doesn't quite manage it.

This is exactly the size of company most of Malaysian manufacturing actually is. It's why the toolkit that solved the last problem doesn't solve this one: an ERP system gives a company one shared version of its numbers, but nothing about it was ever built to notice the moment two departments' pictures of the same order quietly stop matching. A hundred-person manufacturer today typically carries more product variation and a longer, more international supply chain than one its size carried a generation ago — and can't simply hire its way out of the coordination gap, not at the wages and availability of skilled planners in today's market.

Malaysia's own numbers back this up. In manufacturing, AI adoption sits at 50%, but only 13% of manufacturers feel genuinely ready for what's next, and 57% are still stuck in early-stage experimentation. Asked why, the honest answers aren't about ambition: 44% point to technical or data issues, 41% to workforce capability gaps — the exact shape of a business that has the systems but not the layer that would make them talk to each other in real time.

There's a rough industry number worth attaching to this, even loosely. Safety stock — inventory held purely as a buffer, much of it against exactly the kind of information gap described above — typically runs 20 to 30% of its value a year in carrying cost, by widely used supply-chain benchmarks (ASCM, formerly APICS, and research from APQC, the American Productivity and Quality Center). Even a modest slice of a mid-sized manufacturer's buffer stock, sitting there specifically because two departments couldn't fully trust each other's numbers, is a real, recurring cost that no line item calls by its actual name.

The question worth asking before the next planning meeting

Not "do we need AI" — that's the wrong-sized question. The sharper one: right now, somewhere in your own operation, two people are almost certainly looking at two different versions of the same truth. How long does it usually take before that gap surfaces? And what did the last one actually cost — a rush fee, a redone run, a customer who left — that nobody ever added up because it was never given a name?

Sources: Malaysia manufacturing AI adoption and barriers, AWS-commissioned survey via ITbrief Asia; Inventory carrying cost benchmarks, ASCM/APQC data via Eightx; Malaysia MSME performance 2024, DOSM.

← All notes