Before January 2023: How We Ran This Plant on Excel, Paper and Chat Groups

Most buyers never see how an aluminum plant schedules work. They send a drawing, get a price, and later a container. What happens in between is usually invisible — and for most of our history, it was also invisible to us: the plan lived in a spreadsheet, a whiteboard and a group chat.

One note for readers outside China: our internal coordination ran on WeChat groups — that is simply what every Chinese factory uses day to day. Conversations with customers were on email and WhatsApp all along. So the chaos described here is not about the app; it is about having no single source of truth behind it.

This is the first episode in a series about how we digitized our own plant. Not a brochure about Industry 4.0. A record of what we actually did since January 2023, what broke, and what we still do by hand.

What "before" actually looked like

Until January 2023 we had no manufacturing execution system. We did have a standard accounting system for finance and inventory — but it was an island: the shop floor never saw it, and the numbers in it were re-typed or re-confirmed by a person. Everything else was people plus office tools.

Orders. A buyer's drawing came in by email or WhatsApp. The salesperson read it, typed the key data into a spreadsheet, and confirmed by phone. If the customer changed one dimension, the drawing was re-sent and the sheet was edited — sometimes in two copies, and sometimes only one of them was current.

Scheduling. One planner, an Excel sheet with every press and every die, and a whiteboard in the office. Rush orders were inserted by hand. When two departments needed the same press on the same day, the argument was settled by whoever reached the planner first, not by data.

Dies. We hold more than 10,000 dies. Each one had a paper or spreadsheet record: when it was made, which profile it runs, when it was last polished, whether it was nitrided. Finding the right die was easy if the person who made it was at work. If he was not, it could take an hour.

Inspection. First-off, in-process and final checks were written on paper forms and filed by batch. If a customer asked three months later what we measured on his run, someone had to walk to the filing cabinet.

Status. The only reliable way to know where an order was, was to walk to the machine and ask the operator.

The four things that broke most often

None of these are dramatic failures. That is exactly the problem: they are small, constant leaks — half a day here, a scrapped length there — and they never show up on a P&L as a single line.

Why we moved in January 2023

Not because a consultant told us to. Because three things happened at once: order mix moved from mostly architectural profiles to roughly half industrial work (heat sinks, LED profiles, machined housings), deadlines got shorter, and our own people started keeping private spreadsheets to protect themselves. When a plant runs on private spreadsheets, the company has already lost control of its data.

So we implemented a tier-one MES platform alongside that accounting system. Sales orders stayed where they belong — in the ERP. What the MES took over was the shop floor and the plan: production planning and scheduling for extrusion, anodizing, powder coating and machining, plus die and maintenance records, IoT data from the presses, and inspection.

That split is worth spelling out, because it is the opposite of what many ERP vendors promise: in our plant the production plan lives in the MES, not in the ERP. The ERP knows what the customer ordered and what it is worth; the MES decides which press, which die and in which order. In January 2023 five of our eight presses went live on it. The three largest are not on it yet — that is a separate story in this series.

And to be blunt about what a new system does not fix: the systems stayed separate islands. The MES holds the production truth, the accounting system holds the money truth, and the two still do not talk to each other directly. The integration layer is a person — someone exports a report, checks it against the other system, calls the workshop to confirm, and types the result somewhere else. Going digital removed a lot of walking and hunting; it did not remove human confirmation. Anyone who tells you a system alone will fix this has not run a plant.

What we tell other plant owners

• Digitize the work that repeats every day before the work that looks impressive on a slide.
• Write the contract properly: who owns the code, the server and the data. We learned this the hard way.
• Expect your staff to resist, then depend, then worry about their jobs. All three phases are normal.
• Do not assume full automation is the goal. Our scheduling is still system plus human — and it works better that way.

What comes next in this series

• The Excel-and-paper years in detail (this episode)
• One minute to trace a batch back to its billet
• 10,000 dies in one database
• Why we let humans overrule the scheduling system
• The IoT supplier that went bankrupt
• Where your order still stops being paperless: export documents
• What we would do differently

If you are sourcing aluminum profiles and want to know how the plant behind your order actually works, send us your drawing — we will tell you which press it runs on, which die it needs, and how long it will take.

Request a quote