Machine maintenance that does not live on a whiteboard

Breakdowns are reported by word of mouth and tracked on a whiteboard. Spare parts are found by asking around. Here is how a plant gets one list of what is broken, who is on it and which parts are on the shelf, without buying a maintenance suite.

The problem

When a moulding press or an assembly line stops, the operator tells the shift lead, the shift lead finds a technician, and somebody writes it on the whiteboard by the maintenance office. On nights, the whiteboard is often the only record. By the morning meeting nobody is sure how long the machine was down, what was wrong or what was replaced.

Spare parts are worse. The plant has a parts room, but the list of what is in it lives in one technician's spreadsheet, last updated in spring. A heater band that is "definitely on the shelf" turns out to have been used in March, and the line waits two days for a courier.

What was really causing it

The maintenance team is good at fixing machines. What they do not have is a shared memory. Every request travels by voice, every repair is recorded by whoever remembers to, and the parts list depends on one person's discipline.

That has knock-on effects nobody sees directly: the same fault recurs because nobody can look back and spot the pattern, planned maintenance slips because breakdowns always win, and the plant manager cannot tell the owners which machines are costing the most. We confirmed all of this by spending two shifts with the team, including a night, and following every request from the first shout to the machine running again.

The maintenance team is good at fixing machines. What they do not have is a shared memory.

What the research says

The pattern in the NIST data is the one a whiteboard hides: plants that mostly fix things after they break lose far more time and make more defects. You cannot move from reactive to planned maintenance without a record of what broke, why, and what it took, which is exactly what a shouted request and a wiped whiteboard never leave behind.

What we built

A small maintenance app shaped around how this plant already works, not a maintenance suite with three hundred screens. Operators report a problem from a tablet at the line or a QR code on the machine. Technicians see one list, ordered by what is stopping production. Every job records what was wrong, what was done and which parts went in, and the parts list updates itself as parts are used and received.

  1. Reported The operator scans the machine's QR code and picks what is wrong from a short list, or says it in a sentence. No login hunt, no form with twenty fields.
  2. Ranked Anything stopping a line goes to the top. The technician on shift gets a message on their phone straight away.
  3. Fixed The technician records what they found and which parts they used, in under a minute, on the same phone.
  4. Restocked Parts used come off the shelf count. Anything below its minimum becomes a reorder suggestion for the maintenance lead to approve.
  5. Reviewed Each week, one page shows downtime by machine and by cause, so the morning meeting talks about patterns instead of anecdotes.

What stays as it is

The team, the parts room and the purchasing process. Orders for parts still go through the plant's existing purchasing in the ERP; the app only suggests them. The whiteboard can stay on the wall for as long as anyone wants it. In our experience it quietly stops being updated within a month, because the list on the tablet is more useful.

Where AI fits, and where it does not

AI has two small, useful jobs here. It turns an operator's spoken or typed sentence ("press 12 is short-shotting on cavity 4 again") into a structured report: which machine, which part of it, what kind of fault. And once a few months of jobs exist, it can group free-text repair notes into causes, so recurring faults show up.

Everything else is ordinary software: the list, the priorities, the parts counts and the reorder thresholds. Nobody wants a model deciding which line gets a technician first; the rules for that are written down with the maintenance lead and are the same every time.

What it is worth

What comes next

Planned maintenance schedules driven by machine hours instead of the calendar, a running cost per machine for capital decisions, and, if the plant wants it later, machine signals feeding the same list so some faults are reported before an operator notices them.

Questions people ask

Why not buy a full maintenance system (CMMS)?

Sometimes that is the right answer. When the real gap is recording requests, repairs and parts, a small app shaped around how the floor already works is adopted faster, and its records can move into a full system later.

Will operators actually use it?

Only if reporting takes less effort than finding the shift lead. That is why it is a QR code on the machine and a short list, not a form.

Does it connect to our ERP?

Parts orders stay in the ERP. The app suggests reorders, and reads item numbers and stock from the ERP where it allows.

  1. Economics of Manufacturing Machinery Maintenance (NIST AMS 100-34), D. S. Thomas and B. A. Weiss, NIST (2020)

Canonical URL: https://ugibo.com/case-studies/maintenance-off-the-whiteboard