Walk into most small and mid-sized manufacturers and the production plan is on a whiteboard. Job cards live on clipboards and travel with the work. Stock is counted properly once a year and estimated the other three hundred and sixty-four days. And the person who knows what a job really costs is the one who has been quoting them for eleven years.

None of that is incompetence. It is a system that was built by people who were too busy making things to build a better one. It also quietly eats margin, and it never tells you where.

Where the money actually leaks

Manufacturing has an unusual problem: almost everything that determines profitability happens on the floor, and almost all the recording happens in the office, hours or days later. In the gap between the two, detail evaporates.

  • Nobody knows the true cost of a job. Quoted hours are compared to invoiced hours, not to actual hours. Rework gets absorbed. Off-cuts and scrap never make it onto the sheet. The jobs that lose money look exactly like the ones that make it.
  • Work in progress is invisible. Ask what is on the floor right now, in rand, and you get a considered estimate. Your auditors get one too.
  • Machine downtime is folklore. Everyone knows the old press is a problem. Nobody can say it cost you 140 hours last quarter, which is why it never makes the capex list.
  • Traceability is a fire drill. A customer queries a batch from March. If your answer involves a lever-arch file and a good memory, you do not have traceability, you have optimism.
  • Planning is reactive. With no reliable lead times or capacity data, the schedule is built on whoever shouted most recently.

The whiteboard problem

The whiteboard deserves a special mention, because manufacturers love it and it is genuinely good at one thing: everyone can see it. The problem is that it holds one moment in time and no history. When the schedule changes, the previous version is wiped away with a cloth. There is no record of what was promised, what moved, why it moved, or how often it moves. So you can never answer the question that would actually improve the business: what keeps going wrong?

A whiteboard tells you what you plan to do today. A system tells you what you did, what it cost, and what that means for the promise you just made to a customer.

What changes when the floor captures data

The shift is smaller than it sounds. It does not require a fully automated smart factory, and for most South African manufacturers it should not start there. It requires the work to be recorded where the work happens, in a form the system understands. A tablet or scanner at each work centre. A job that gets booked on and booked off. Material issued against a job instead of quietly disappearing from the store.

From that one change, a surprising amount follows. Real job costing, so quoting becomes evidence-based instead of instinctive. Live work in progress, so month-end stops being archaeology. Actual cycle times, so promised delivery dates start being met. Genuine batch and serial traceability, so a recall is a query rather than a crisis. And a maintenance history, so you can argue for a new machine with numbers instead of frustration.

Start with the job card, not the ERP

The single highest-return move for most manufacturers is digitising the job card: who worked on what, for how long, using which materials, with what result. It is one form, it takes weeks not quarters, and it produces the data every other improvement depends on. Full ERP, if you ever need it, is far easier to justify and to implement once that habit exists.

The people part, honestly

The floor will not adopt anything that makes their day slower. This is the reason most shop-floor systems fail, and it is entirely predictable. If booking onto a job takes six taps and a login, it will be done at the end of the shift, from memory, badly, which is exactly the problem you started with, now with a licence fee attached.

The test is brutal and simple: capturing the data must take less effort than the paperwork it replaces. Scan a badge, scan a job, done. Give the supervisors the screen that saves them their morning walk-around. Show the team the numbers they helped create, so it is a tool rather than surveillance. Get that right and adoption looks after itself.

What good looks like a year in

You can price a job by pulling up the last three of its kind and seeing what they truly cost. You know what is on the floor, in rand, this morning. You can tell a customer their delivery date with a straight face. You can trace a batch in minutes. And the knowledge that used to live entirely in your longest-serving estimator now lives in something that does not resign.

The machines were never the hard part. Knowing what they actually did all week is.