Automating OEE usually makes the number worse.
Manual logging cannot see stops under about fifteen minutes. When those reappear as measured downtime, availability drops — and that drop is the most valuable output of the first fortnight.
Most Indian plants that track OEE do it manually, from a logbook, compiled at shift end. That method is not worthless — it is how OEE was done for thirty years and it produces a usable trend. But it has a structural blind spot, and knowing exactly where that blind spot is tells you what automating would actually buy.
It also explains why a plant's OEE almost always falls when it automates, which catches management off guard often enough to be worth warning about in advance.
| Manual | Automated | |
|---|---|---|
| Shortest stop recorded | Roughly 15 minutes, in practice | Seconds |
| Who records it | An operator or supervisor, from memory | The machine, as a by-product of running |
| When it is available | Days later, after compilation | During the shift it describes |
| Cost per shift | Real: 15–30 minutes of someone's time | None, after installation |
| Consistency | Varies by person and by how busy the shift was | Identical every shift |
| Disputability | High — everyone has a version | Low — one shared record |
| Reason quality | Often better, because a human was there | Depends entirely on reason capture design |
The last row is worth noting, because it is the one place manual can win. A supervisor who was present writes a richer reason than a two-tap menu selection. Good automated systems close this by making reason entry fast enough to be honest rather than exhaustive.
Manual logging cannot see short stops, so they are absent from the loss record and the time simply vanishes into the shift. When automatic collection begins, those minutes reappear as measured downtime, and availability drops.
The drop is often substantial — a plant that believed it was at 70% commonly lands in the fifties on first honest measurement. Nothing got worse. The difference between the two figures is a reasonable estimate of what was never being counted, and it is the most valuable single output of the first fortnight.
Management that is not warned about this concludes the system is faulty, or that someone has been misreporting. Both conclusions are wrong and both are damaging. Say it in advance, in writing, and the drop becomes the first finding rather than the first argument.
A handful of machines, long cycles
On four machines with two-hour cycles, manual logging captures nearly everything that matters and costs very little.
Proving the concept
A fortnight of manual measurement before buying anything is a legitimate and cheap way to establish whether a loss problem exists at all.
Short cycles or many machines
Above roughly ten machines, or on cycles under a few minutes, manual logging loses more than it captures.
Where the number is disputed
One shared record ends an argument that a logbook cannot.
Manual OEE is not free. Fifteen to thirty minutes per shift of supervisory time, plus compilation, plus the meetings spent disputing the result. Across a two-shift operation that is a meaningful recurring cost, and it buys a number available days late at fifteen-minute resolution.
Automation carries an upfront cost and then produces a better number at no recurring labour cost. Whether that trade works depends on your machine count and what the losses are worth, which is arithmetic you can do yourself with the downtime cost calculator and the ROI calculator. If the recoverable loss is not comfortably larger than the cost, the honest answer is to stay manual — and you will have established that for free.