Busy machines and missed dispatch are not a contradiction.
Production monitoring measures output against intent, not capacity. What it tracks, where the data comes from, and why the board belongs on the floor rather than in a report.
| Hour | Plan | Actual | Gap |
|---|---|---|---|
| 06–07 | 60 | 58 | −2 |
| 07–08 | 60 | 61 | +1 |
| 08–09 | 60 | 44 | −16 |
| 09–10 | 60 | 39 | −21 |
| 10–11 | 60 | 57 | −3 |
The gap opened at 08:00 and was on the floor board while there was still time to close it, not in tomorrow's report.
Machine monitoring and production monitoring answer different questions, and plants that conflate them end up with dashboards nobody in planning uses. Machine monitoring asks whether a machine was running. Production monitoring asks whether the plan was met — which involves the machine, but also the sequence, the material, the operator and the schedule.
The distinction matters commercially. A machining floor can post excellent machine utilisation and still miss dispatch, because the wrong parts were made in the wrong order. Utilisation is a machine metric; adherence is a production metric.
Output against intent, continuously.
Hour by hour, not shift by shift
A shift-end comparison tells you that you missed. An hourly comparison tells you when it started going wrong, which is the only version anyone can act on.
Right parts, right order
Whether the sequence actually run matched the sequence planned. Deviations are usually rational at the time and invisible afterwards.
Where each batch is
Progress per route card from real part counts, so the answer to where an order stands takes ten seconds rather than a walk.
Good output, not gross output
Production that has to be reworked is not production. Rejections logged at the machine keep the count honest.
Counts, states and one human input.
Part counts come from the machine — from the control's own counter on a CNC, from a stroke counter or proximity switch on a press, from a cycle-completion event on an SPM. This is the spine of production monitoring and it is the part that must not depend on a person, because manual counting fails exactly when the shift is busiest.
Machine states supply the context: a shortfall against plan looks entirely different if the machine was running the whole time than if it was stopped for two hours. Without state data, an hourly board shows the problem and hides the cause.
The one input worth asking a human for is the job or route card selection, because a machine cannot know which order it is working on. On MachineWise floors this is a scan or a two-tap selection at the kiosk, and everything downstream — batch progress, per-order OEE, traceability — depends on it.
The board, and why it belongs on the floor.
An hourly plan-versus-actual board on a TV screen at the end of the line changes behaviour in a way that a report to management does not. People adjust when they can see the gap while there is still time to close it, and that is the entire mechanism — visibility early enough to act.
The screens run themselves: auto-login, auto-loop, auto-recover after a power cut. That sounds like a detail and is not. Shop-floor displays that need someone to log in and click are dark within a fortnight, and once a screen has been dark for a week nobody looks at it again even after it is fixed.
For supervisors who are not on the floor, the same data arrives as a shift summary on WhatsApp. For planning, it arrives as schedule adherence and a live WIP position rather than a spreadsheet compiled the next morning.
Good utilisation, missed dispatch.
A components unit running two shifts posts machine utilisation in the mid-seventies and consistently misses weekly dispatch. Machine data alone offers no explanation, because by that measure the plant is doing well.
Adding job selection and hourly plan-versus-actual shows the sequence: high-volume parts run first because they are easy and keep utilisation high, while the low-volume items with the nearest delivery dates queue behind them. The machines were busy; they were busy on the wrong work. The correction is a scheduling rule, not a capital purchase — and it was invisible until output was measured against intent rather than against capacity.
Put a plan-versus-actual board on your line.
We connect the machines, set up job selection at the kiosk and run the hourly board on a TV at the end of the line. Two machines, thirty days, free.