Here is the method. It is manual on purpose, it costs nothing, and it produces a number that survives an audit. Automate later if you want. Count first.
One asset, one sheet
A single tally sheet at the constraint. Three columns: time, duration, category. Nothing else.
The temptation is to add fields. Shift, operator, product, root cause, work order number. Resist all of it. Every column you add lowers the completion rate, and a partially filled sheet with eight columns is worth less than a completely filled sheet with three.
Five categories, matched to what we covered Monday:
- Unplanned stop or breakdown
- Running below rate
- Changeover or startup
- Minor stop or short interruption
- Quality, rework, or hold
Count the small ones
This is the instruction that has to be explicit, because it runs against everything the crew has been trained to do.
Anything that interrupts flow gets a mark. Including the ninety second clear. Including the jam that one person fixed without telling anybody. Including the reach in to reposition a case that the operator has done so many times it no longer registers as an event.
These are the ones your system has never seen, and they are the reason the counting exercise exists. If the sheet only captures what the system already captures, you have spent two weeks confirming what you knew.
Log the rate
Once an hour, record actual output against design. One line, one number.
Rate loss is invisible without this, because nothing stops and nothing alarms. It is also, in a large share of plants, the second largest category on the sheet. Skipping this line is the single most common reason a counting exercise underestimates the problem.
Reconcile weekly
This is the step that separates a defensible number from a tally exercise, and it is the step nearly everyone skips.
At the end of each week, total your counted hours and compare against the gap you calculated in week two. Those two numbers were derived completely independently, one from a design rate calculation and one from direct observation, so they act as a check on each other.
If your counted hours land within roughly ten percent of your calculated gap, your number is defensible and you should stop worrying about it.
If they do not, the difference is telling you something specific. Counted hours well below the calculated gap usually means the counting is missing a category, and it is almost always rate loss or minor stops. Counted hours well above the gap usually means the design rate is too high, or scheduled hours were overstated.
Either way you have learned something before you present, rather than in front of the person deciding whether to fund you.
Write your prediction first
Before the first shift, write down what you expect to find. Which category will be largest, roughly what share.
Then count anyway.
The gap between the prediction and the result is where the credibility comes from. Being wrong in a documented way is far more persuasive to a skeptical executive than being right by assertion, because it demonstrates that the process produced the answer rather than confirming a position you already held.
Two weeks, not one
Two weeks is the minimum that spans enough changeovers, product runs, shift patterns, and crews to be representative.
One week produces a number a skeptic can dismiss in a sentence, and they will, because dismissing it is easier than acting on it. Do not hand them the sentence.
