A part is in the good count only if it meets the spec of the order being run, the first time, with no rework, at one fixed counting point. Scrap, rework, destroyed samples and downgrades are out, and held stock stays out until it is released.
Why a small rule decides a big number
The good count is the one number that drives OEE directly. OEE = good count × ideal cycle time ÷ planned time, so each unit you wrongly add goes straight into the result.
The arguments are rarely about scrap. They are about the grey cases, like a part fixed on the spot or a bottle cut open by the lab.
The rulings, case by case
| Case | Good? | Why |
|---|---|---|
| Passed first time, in spec | Yes | This is the definition |
| Scrapped | No | Waste |
| Reworked, then passed | No | Not right the first time |
| Destroyed in a test | No | Testing waste, cannot be sold |
| Checked, then returned to the line | Yes | Nothing happened to it |
| Downgraded to second grade | No | Missed the spec of this order |
| Made to a lower-grade order, and in spec | Yes | It met the order's spec |
| On hold for a result | Not yet | Good once released |
| Start-up unit, in spec | Yes | No rule says early units are bad |
| Found bad after shipping | Move to waste | Log it as waste, with the cause |
What each ruling does to OEE
Example: a blow moulding line has 480 scheduled minutes, an ideal rate of 50 bottles per minute, and makes 20,000 bottles. Of those, 400 are scrap, 300 are reworked, 240 are downgraded and 60 are cut open by the lab.
Now start with the most generous rule, and tighten it one step at a time.
The right answer is the fourth bar: 19,000 ÷ 50 = 380 good minutes, and 380 ÷ 480 = 79.2%. Quality reads 19,000 ÷ 20,000 = 95.0%, where the most generous rule would say 98.0%.
The last bar shows a shift where 500 of those good bottles were still waiting on a lab result at the time of the report. They read as not good for now, and they count when released.
Rework: the most argued case
A unit that needed rework was not right first time, so it is out, even if it later ships at full price. That is the rule in Overall Equipment Effectiveness (R. C. Hansen, 2001), and it treats the original time as lost.
A simple touch-up is still rework. If you find this rule hard to accept, remember that the time and the cost were real, and only a count that shows them gets them fixed.
How to treat the rework run itself is covered in how to calculate OEE quality. For the plant-wide view, see how to reduce scrap and rework.
Samples: ask what happened to the part
A part destroyed by a test cannot be sold, so it is testing waste. A part measured and put back is still good, since testing it changed nothing.
Retained samples kept in a library and never sold are out as well. Our advice is to write the planned sample count per run into the standard, so nobody argues about it each shift.
A sample that is standard for every run still counts as a loss, even though everyone knows it is coming. Waste that shows up every run tells you about the design of the process.
Downgrades: judge against the order
If the order is for first grade and a unit ends up second grade, it did not meet the spec. It is out, even if a customer will pay for it.
A second-grade order is different. If the plan said second grade and the unit meets that spec, it counts as good, and OEE does not judge the price.
Keep the money view apart. Cost of poor quality shows what a downgrade really costs, since OEE only shows time.
Sort every unit in four questions
A downgrade fails here.
Rework fails here.
Destroyed and retained samples fail here.
Held stock waits here, and counts on release.
Where to count, and how to keep it honest
Pick one counting point and keep it fixed. For a line, that is normally at the end of the step or at transfer, after any reject station.
- Total and good from one window. Take both counts for the same shift or run, or quality can read over 100%.
- Good = total − not good. Or count good and rejects apart, then check they add up to the total.
- Reconcile with packed units. Compare the good count with pallets or cases at least once a week.
- Late finds go in as waste. Do not quietly rewrite last week. Write it where it was found, with the cause.
The options for hardware are in part counting sensors. If the good count looks too large, see OEE over 100 percent.
Good count vs similar measures
- First pass yield uses the same first-time idea, usually for one process step.
- First time through applies it across a process or a value stream.
- Rolled throughput yield multiplies it across steps.
- Scrap rate counts only what is thrown away, so rework and downgrades do not show. That makes it look better.
So a plant can have a 2% scrap rate and a 95% quality factor on the same bottles, as in the example, without either being wrong. The first pass yield calculator is quick for this kind of check.
Counting errors to rule out
- Counting good units at the machine and bad units at final inspection, in different periods.
- Calling a run all good because the area passes everything on with no check. Adjust with your history instead.
- Different crews using different rules for touch-ups.
- Using good units in performance, which counts the same bad units twice. See OEE calculation mistakes.
A good count holds up best when total units come from the machine and not from a shift sheet. Fabrico is a light MES for OEE that takes counts, speed and stops live from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, with OEE per line and shift. See how Fabrico counts per line.
What people ask next
Does rework count as good in OEE?
No. A reworked unit was not right the first time, so it stays out of the good count even if it ships at full price. See how to calculate OEE quality for how the rework run is counted.
Do start-up units count as good?
Yes, if they meet the spec on the first pass. Start-up units that fail are waste, like any others, and startup losses shows how to track them.
Are quality samples counted in OEE?
Destroyed or retained samples are out of the good count, since they can never be sold. Parts checked and put back stay in.
Should total count include scrap?
Yes. Total count is everything the machine made, good and bad, and quality is good count ÷ total count. If scrap is missing from the total, quality reads too high and performance too low.
What if a good part is damaged later, at another step?
Count it where it happened. It was good at this step, so it stays in this step's count, and the other step records the loss. Compare with production yield for the whole-process view.
Fabrico is a light MES for OEE. It takes stops, speed and counts from PLCs or IoT sensors, with add-on AI cameras counting output where a machine has no usable signal, and operators tag the cause at the line.