Good Count vs Total Count: The Two Numbers OEE Needs

Total count is every unit a process produces in a run, good and bad together. Good count is the subset that passed the first time, with no rework and no scrap. Every OEE calculation needs both: the Quality factor is good count divided by total count, and the gap between them is your defect loss.

These two numbers look trivial until you try to collect them honestly. The traps are all in the definitions: what counts as "good," where on the line you count, and how a reworked part gets tallied. Get any of those wrong and Quality reads high while scrap piles up in the corner. This guide pins down each count, shows how counting at the wrong place double-counts reworked parts, and connects the pair to OEE and first pass yield.

What is the difference between good count and total count?

Total count is the gross number of units the process ran; good count is the number that came out right the first time. Total count includes everything the machine cycled, sellable units, scrap, and any unit that needed a second pass to fix. Good count strips out the scrap and the rework, leaving only first-pass-good units. The difference, total minus good, is the reject count: everything that failed to be right the first time.

The phrase "the first time" is the whole game. A unit that was defective, got reworked, and now passes is not good count. It consumed extra time and effort, hid a process problem, and only became sellable through a recovery step. Counting it as good erases the signal that something went wrong. Good count is deliberately strict because its job is to make defects visible, not to measure what you can eventually ship.

Why does good count feed the OEE Quality factor?

Because the Quality factor exists to convert defect loss into time, and it does that as good count divided by total count. In OEE, Quality = good count ÷ total count. A run of 10,000 units with 9,400 first-pass-good gives a Quality of 94.0%. That percentage then multiplies Availability and Performance to give OEE so every reworked or scrapped unit drags the whole number down, which is the point.

There is a neat cross-check hiding here. OEE can also be written as good count × ideal cycle time ÷ planned production time. That form uses good count directly, the time spent making first-pass-good units, as a fraction of the time you planned to run. If Quality counts rework as good, this shortcut and the three-factor version stop agreeing, and a disagreement between the two is a reliable sign that a count is wrong. The two counts are also the raw material for the six big losses: total minus good splits into startup rejects and steady-state production rejects, the two quality losses.

Where should you count, and why does place matter?

Count at one fixed point per run, and count reworked parts only once. The classic error is counting in two places, say, at the machine's cycle counter and again at a downstream inspection or repack station, so a part that gets pulled, fixed, and sent back through is cycled twice and shows up twice in total count. Now total count is inflated, Quality is distorted, and the numbers stop tying to reality.

The safe rule is to define total count as units the process attempted once, and good count as units that cleared inspection the first time, both read at the same location. If a reworked part re-enters the line, it must not add to total count again, it already counted on its first pass. Machine cycle counters are usually the cleanest source because they count attempts at one physical point, but they need a matching quality read so you know how many of those attempts came out good. When counts are tallied by hand at several stations, the double-count creeps back in; that is a big part of why counting close to the machine, at a single defined point, beats reconstructing counts from paperwork later.

How do reworked parts get miscounted?

Reworked parts get miscounted in two opposite directions, and both distort Quality:

• Double-counting the unit. The part cycles once, fails, gets reworked, and cycles again through the same counter, so total count goes up by two for one physical part. This understates Quality and overstates output. The fix is to count attempts at a single point and not re-add a returning part.
• Counting rework as good. The reworked part finally passes, and someone records it in good count because it is now sellable. This overstates Quality and hides the defect entirely, the run looks like it made 9,700 good when it really made 9,400 first-pass and rescued 300. The fix is the strict definition: good means right the first time, full stop.

The two errors can even cancel out on the top-line units-shipped number while leaving Quality meaningless, which is the worst case, the plant looks fine and has no idea its first-pass rate is slipping. The discipline that prevents both is to record three numbers per run, not two: total attempts, first-pass good, and rework recoveries. Good count is the middle one, and it never includes the third.

How do you set up the two counts on a line?

Getting trustworthy counts is a setup problem you solve once. Here is the sequence:

1. Define total count as attempts at one point. Pick a single physical location, usually a machine cycle counter, and declare that its count is total count. Write it down so nobody adds a second source later.
2. Define good count as first-pass passes. A unit is good only if it clears inspection without rework. Put the quality read at or right after the same point you count total.
3. Decide the rework rule explicitly. State in writing that a reworked part does not re-increment total count and never counts as good. Track recoveries in a separate tally.
4. Match the counts to the same time window. Both counts must cover the same run and the same period, or Quality is comparing mismatched populations.
5. Reconcile against units shipped. Periodically check that good count plus rework recoveries minus later failures ties to what actually shipped. A persistent gap means a count is leaking somewhere.

What do the standards say about the two counts?

The definitions are not folklore, they are written into the international standard for manufacturing KPIs.

ISO 22400-2:2014 the international standard for manufacturing-operations KPIs, defines produced quantity and good quantity and specifies the quality ratio as good over produced, the same structure OEE uses (standard listing). The standard exists precisely so that "good" and "total" mean the same thing from one plant or vendor to the next, which is why pinning the definitions down locally matters: a Quality number is only comparable if everyone counts the same way. Treat the standard as the tiebreaker when someone argues a reworked part should count as good, by the definition, it does not.

How do the two counts connect to yield and OEE?

Good count and total count are where quality metrics start branching. The Quality factor in OEE uses them directly; first pass yield is essentially the same ratio expressed per station, and rolled throughput yield multiplies first-pass yields across stations to show how defects compound down a line. Where OEE's Quality is a single-station or single-process view, FPY and RTY extend the same good-over-total logic across the whole flow, so a line can post a decent per-station Quality and still have a poor rolled yield once every station's first-pass rate is chained together.

The practical payoff of getting the counts right is that every downstream metric inherits their honesty. Soft counts, rework smuggled into good, parts double-counted, quietly corrupt Quality, OEE, yield, and scrap cost all at once. Counting attempts and first-pass passes at a single point, from machine signals rather than end-of-shift memory, is the fix, and it is the same source data Harmony uses to compute true OEE on the floor (see the platform), shown in the CLS case study. There is also a throughput angle: only good count turns into shippable output, so a rising total count with a flat good count is pure motion without progress, the reason throughput in manufacturing is defined in good units, not gross cycles. And a spike in reject count often traces back to the same stops and speed losses tracked in machine downtime since a struggling machine tends to make bad parts before it makes none. Run your own good and total counts through the OEE calculator and read manufacturing KPIs for how Quality sits alongside the other numbers on the board.