Parts production: four stages at 96% give 85.78%

“The report says average quality of ninety-six and a quarter percent. I launch twenty-two thousand blanks, and eighteen thousand eight hundred and seventy-two parts arrive at the warehouse. For a year and a half I could not work out which of those two numbers was lying” — that is how Ihor, owner of a plastic parts production in Zaporizhzhia, opened the conversation.

Sound familiar? The moulds are run in, moulding is stable, there are almost no complaints. Every month the inspector files a report in which all four sections hold their percentage. And the profit looks as if the shop were working twice as badly as its own figures.

The cause is neither quality nor people. It is the arithmetic everybody uses and almost nobody checks: the quality of four consecutive sections is counted as the average across sections. And out comes a number that has nothing to do with how many parts left the shop.

We break production down into two units: the stage and the part that passes through it. And we show why the same defect costs 46 ₴ or 117 ₴ depending on where it was caught.

The founder's path: from one machine to four sections

Ihor started with a single injection moulding machine bought second-hand. He moulded housings for a local assembler and stood at the machine himself. Scrap was physically visible back then: a short-shot part flew into a crate, and the crate stood by his foot.

Over seven years a second and a third machine were added, a milling section, a paint booth with a drying oven and an assembly section. The launch grew to twenty-two thousand blanks a month.

“When there was one machine, I knew everything about scrap, because it lay at my feet. Now there are four sections and each has its own crate. Percentages make it into the report, not crates.”

Why “average quality of 96.25%” says nothing about yield

The percentages at the sections are honest: moulding holds 96%, machining 97%, painting 94%, assembly 98%. Every figure is verifiable and deceives nobody.

The deception is in the next line, where those four percentages are added up and divided by four. Out comes 96.25% — and that number starts living as the shop's indicator. But 96.25% is the arithmetic mean across stages, the success rate of one abstract section, not the share of blanks that reached the finished-goods store.

What Ihor's report said

The report was short: launch 22 000 blanks, average quality across the four sections 96.25%, price of a finished part 148 ₴. From that a conclusion grows by itself: scrap takes about four percent, that is 825 parts.

In reality 18 872 parts reached the store and 3 128 did not — four times more than the report promised. And not one of its lines was untrue.

Losses do not add up — they multiply

Moulding loses 4%, machining 3%, painting 6%, assembly 2%. The temptation to add them into 15% and say “there, that is the whole loss” is almost irresistible.

You cannot count it that way. Each section works not with the full launch but with what the previous one handed over: machining loses its 3% not from 22 000 but from 21 120; painting its 6% — from 20 486. Losses do not add up, they multiply: 1 − 0.96 × 0.97 × 0.94 × 0.98 = 14.22%, and the shop's yield is 85.78%.

15% against 14.22%: where the trifle is and where the chasm

Between adding and multiplying the difference is laughable — less than a percentage point, and that is exactly why the multiplication is ignored. But the chasm is not here. It is between 96.25% and 85.78%: the mean divides the losses by four, while consecutive sections accumulate them. The more stages in the chain, the further the mean drifts from the true yield — and across four sections it drifts by 10.47 percentage points. The money lies in that gap.

“People explained it to me on their fingers and I did not believe them. I believed it when I recounted by the physical crates: there were three thousand one hundred and twenty-eight parts in them, not eight hundred.”

Life before the stage became a line

The bookkeeping Ihor lived in for almost five years will be recognised by any production that grew out of a single machine.

• Yield was counted as the average across sections. Four percentages were added up and divided by four — and not once multiplied.
• Scrap was measured in pieces, not in money. “Painting has more than moulding” — with no idea how much each of those pieces already costs the shop.
• All sections weighed the same. The percentage at assembly and the percentage at painting were read as figures of equal weight.
• Raw material was written off by output, not by input. The costing carried as much material as sits in a finished part, while the whole launch was paid for.
• Regrinding was treated as money coming back. “It all goes back into the hopper anyway” — with no distinction between what actually goes there and what no longer does.
• Scrap had no address. The crates stood apart, and the report carried one consolidated waste figure.

None of these six problems is technological. All six are accounting ones — and that is exactly why neither the foreman nor the inspector saw them.

The true yield: 18 872 parts out of 22 000 blanks

Ihor laid the month out by stages: how much came in, how much the section rejected, how much a part costs at the entrance of that very section and how much money sits in its crate.

Stage	Stage yield	In, pcs	Scrap, pcs	Accumulated unit cost at entry	Cost of scrap
Moulding	96%	22 000	880	46 ₴	40 480 ₴
Machining	97%	21 120	634	64 ₴	40 576 ₴
Painting	94%	20 486	1 229	86 ₴	105 694 ₴
Assembly	98%	19 257	385	117 ₴	45 045 ₴
Finished parts	85.78%	—	3 128	—	231 795 ₴

The chain reads row by row: 22 000 at the entrance to moulding, 21 120 at the entrance to machining, 20 486 at the entrance to painting, 19 257 at the entrance to assembly, 18 872 in the finished-goods store. The revenue from them is 18 872 × 148 ₴, that is 2 793 056 ₴. And 231 795 ₴ settles in the four crates — money already spent on parts that never went to the customer.

Scrap gets more expensive with every stage

Now the second turn, and it matters more than the first. A part rejected at moulding costs the shop 46 ₴ — raw material only. The same part pulled at assembly costs 117 ₴, because it carries milling, paint and drying on it. The difference is 2.5 times over, and it is invisible in any quality percentage.

Why 86 ₴ is not the price of painting

It is easy to get confused here. The figures 46, 64, 86 and 117 ₴ are not the prices of the stages, nor what it costs to paint a part. They are the accumulated cost of one part at the moment it enters a section.

At the entrance to painting a part already has 86 ₴ sitting in it: raw material plus moulding with machining. If the booth rejects it, the shop loses the whole 86 ₴, not the cost of the paint. At the entrance to assembly 117 ₴ sits in it. Hence the rule for any industry: control at the start of the chain is cheaper than control at the end.

Painting: not the worst percentage and the most expensive scrap

Painting has a yield of 94% — no catastrophe, a working percentage for a booth with a drying oven. In the report it looked like one of four normal figures. But in money it produces 105 694 ₴ of scrap, that is 45.6% of the shop's entire cost of scrap.

There are two reasons, and they act together. Painting rejects the most pieces — 1 229 against 880 at moulding and 385 at assembly. And each of them weighs 86 ₴, not 46 ₴. Multiply them out, and you get the most expensive crate in the shop.

“For two years I was about to change the moulding machine, because moulding has a lot of scrap. And all that time the money was burning in the booth, the one with a normal percentage in the report.”

What this is in the shop's money

To understand the weight of 231 795 ₴, let us put it next to the shop's month.

Cost item	Per month	% of revenue
Raw material (22 000 × 46 ₴, paid at input)	1 012 000 ₴	36.23%
Workers	812 000 ₴	29.07%
Electricity	284 000 ₴	10.17%
Moulds and tooling	96 000 ₴	3.44%
Paint and chemicals	142 000 ₴	5.08%
Other	86 000 ₴	3.08%
Total	2 432 000 ₴	87.07%

Revenue of 2 793 056 ₴ minus costs of 2 432 000 ₴ gives a profit of 361 056 ₴ and a margin of 12.93%. For a plastic parts production that is normal, and Ihor was happy with that figure.

231 795 ₴ is 64.2% of the profit, and it already sits inside the costs

The cost of scrap amounts to 64.2% of what the shop earns in a month. And here a clarification is needed: these 231 795 ₴ are not added to the costs and not subtracted from the profit.

They are already inside the 2 432 000 ₴: the raw material for the rejected parts has been bought, the wages for processing them have been paid, the electricity for moulding them has been burned. This is not a new hole but the part of the existing costs that did not become product. The question sounds different: what share of its costs does the shop pay for something that will never go to a customer.

Hidden item one: raw material is paid for at the input

What was missing from the costing stands in the very first cost line: the shop buys raw material for the whole launch, not for what came out. At the input, 22 000 blanks at 46 ₴ — 1 012 000 ₴. In the finished goods there is raw material worth 18 872 × 46 ₴, that is 868 112 ₴. The difference is 143 888 ₴, or 39.9% of the profit.

This is not a separate hole on top of the 231 795 ₴ — quite the opposite: 143 888 ₴ is the raw material part of those same 231 795 ₴, their largest component. The rest is made up of wages, electricity, paint and tooling. These two numbers must not be added together — it is the same money seen from two sides.

Hidden item two: half the scrap cannot be reground

The owner's firmest belief is that scrap in plastics is almost free, because it is ground up and returned to the hopper. Partly true: moulding and machining give 1 514 parts that really can be ground. The return is not complete — recycled material loses its properties, and about 60% makes it back into work. So 1 514 × 46 ₴ × 0.6 is 41 786 ₴, or 11.6% of the profit, which the shop wins back. And painting with assembly give 1 614 parts that nothing can reground: they carry paint, and ground plastic will leave inclusions in the next batch.

Why 1 614 parts return nothing

Let us label every number. 1 514 is the regrindable scrap, from moulding and machining, and it returns 60% of the raw material. 1 614 is the non-regrindable scrap, from painting and assembly, and it returns zero.

The difference between them is one digit, and in money it is fundamental. The regrindable ones are the cheapest parts in the chain, at 46 and 64 ₴. The non-regrindable ones are the most expensive, at 86 and 117 ₴. In other words, what comes back is the part of the scrap that costs the least.

“I used to think our scrap was cheap, because it all goes into the grinder. Until it hit me: the cheap half goes into the grinder and the expensive half goes to the landfill.”

The average across stages against the true yield

Let us put both figures into one table and look at what each of them promised.

How the yield was counted	Indicator	Finished parts	In money
Arithmetic mean of the four stages — that is what the report said	96.25%	21 175	3 133 900 ₴ of revenue
Multiplication of the stages — the true yield	85.78%	18 872	2 793 056 ₴ of revenue
Gap	10.47 pp	2 303 parts	340 844 ₴
Scenario: painting at 97% instead of 94%	88.52%	19 474	2 882 152 ₴ of revenue

The first row is not the inspector's mistake but a consequence of the formula: the average across stages of 96.25% promised 21 175 parts and revenue of 3 133 900 ₴. The second is what there is. Between them lie 2 303 parts and 340 844 ₴ of revenue that never existed: those parts lived only inside a wrongly assembled average.

And one last thing. 96.25% cannot be “corrected” into 85.78% — these are two different indicators: the first describes an average section, the second the shop. Only the second is any good for planning a launch.

Three percentage points at painting — 89 096 ₴

Multiplication works the other way round too: if losses accumulate, then an improvement at one section pulls the whole chain up. Ihor took the most expensive crate and worked out what happens if the painting yield is raised from 94% to 97% — to the level of the machining next door. The chain recalculates like this: moulding 22 000 → 21 120, machining → 20 486, painting → 19 871 instead of 19 257, assembly → 19 474 instead of 18 872. The overall yield is 0.96 × 0.97 × 0.97 × 0.98 = 88.52%.

Why this gain is almost clean

The gain is 602 parts. At a price of 148 ₴ that is 89 096 ₴ of revenue, or 24.7% of the profit. And there are almost no new costs inside those 89 096 ₴: the raw material has already been bought for the whole launch, the workers are already on shift, the electricity for moulding and machining those parts has already been spent. Three percentage points at one section turn costs already incurred into product.

Compare that with the other way of getting 602 parts — increasing the launch. Then you have to buy more raw material, pay more for moulding and machining, load the machines: the same result, but at full cost.

Why scrap is not visible in the report

If all of this is so obvious in the calculation, why does the average across stages stand in the reports of half the production shops out there? The reasons are not mathematical.

• A section's percentage is easy to measure. The inspector sees his own section and files his own figure, and nobody is obliged to multiply all four.
• An average sounds solid. 96.25% looks like a serious factory number, while 85.78% looks like grounds for an inquiry.
• Scrap has no price. As long as the crate holds “pieces” rather than 86 ₴ apiece, the painting crate and the moulding crate look comparable.
• The grinder creates a feeling of return. It seems the money goes round in a circle, though in reality 60% of the cheapest half comes back.
• Raw material is closed off in a single line. The report carries 1 012 000 ₴ of purchases, and nowhere does it say that 868 112 ₴ of it sits in the finished parts.

Why “tighten up the control” does not help

The owner's first reaction is always the same: add an inspector and fine people for scrap. Ihor did exactly that, and the percentages did not budge.

Because an inspector without prices at the stages has nothing to steer by. To him it makes no difference where a part is caught — at moulding for 46 ₴ or at assembly for 117 ₴: one rejected piece will appear in the report either way. As long as a stage has no unit cost at entry, early and late scrap weigh the same.

“I fined people for scrap for a year and a half. What I should have done was write on every crate how much money is in it. After that the foremen moved the check to the painting entrance themselves.”

What changed once the stage became a line

Ihor did three things. The shop's yield began to be counted by multiplying the stages rather than averaging them. Every section got an accumulated unit cost at entry. And scrap from the crates began to be entered into the books in hryvnias, regrindable and non-regrindable separately. Here is what became visible over two quarters:

• Launch planning stopped lying. The volume for a customer is counted from 85.78%, not from 96.25%, and the shop stopped promising parts that will not exist.
• The control moved to the painting entrance. Pulling a part before the booth, while it holds 86 ₴, turned out cheaper than hunting for a defect in a painted one worth 117 ₴.
• Painting got a resource of its own. 602 parts and 89 096 ₴ with almost no new costs — the cheapest revenue the shop can buy.
• Non-regrindable scrap stopped being invisible. The 1 614 parts in which everything invested has burned are now a line of their own, and the 1 514 regrindable ones another.

None of these decisions required new equipment or new people. They required seeing the stage as a line — with a yield of its own and a unit cost of its own at entry.

The takeaway for business owners. Wherever a product passes through several consecutive stages, quality is not averaged — it is multiplied. The average percentage across sections always looks better than the true yield, and the more stages there are, the wider the gap. And the price of scrap depends not on the defect but on how many stages the part has already passed: pulling it at the entrance is always several times cheaper than at the exit. It works the same way in metalworking, in a sewing shop and in a food production — you only have to label each stage with how much money is already lying in it.

Parting advice

• Count the yield by multiplying, not by averaging. 0.96 × 0.97 × 0.94 × 0.98 gives 85.78%, while the average across the same stages shows 96.25%.
• Put an accumulated unit cost at the entrance of every section. 46, 64, 86 and 117 ₴ are not the prices of the stages but how much a part already costs the shop when it enters a section.
• Convert the scrap crates into hryvnias. In pieces the biggest crate may be the cheapest: 1 229 parts at painting give 105 694 ₴, while 880 parts at moulding give 40 480 ₴.
• Look not for the worst percentage but for the most expensive crate. For Ihor the two coincided at painting with its 45.6% of the entire cost of scrap, but they do not always coincide.
• Compare the raw material purchased with the raw material in the finished goods. 1 012 000 ₴ at the input against 868 112 ₴ in the product gives 143 888 ₴ — the fastest way to see the losses.
• Split scrap into regrindable and non-regrindable. Out of 1 514 parts from moulding and machining 60% of the raw material comes back; out of 1 614 after the paint, nothing does.
• Count the gain in yield, not the gain in volume. Three percentage points at one section gave 602 parts on raw material already bought.

The money does not disappear — you simply do not see it

A plastic parts production looks like a business where everything is decided by the moulds, the granulate and the machine settings. In reality that is the ticket to entry, and the money is decided by something else: how many blanks out of the launch reach the finished-goods store, and at which stage you pull the rest.

This gap will not close by itself — the averaging formula will carry on showing a pretty number. But the moment every stage gets a yield of its own and a unit cost of its own at entry, you can see where the money burns, which crate is worth the effort this week — and how many parts you are paying full cost for just so that they can make it to the bin.

How to see this in your own production

Everything described above rests on one kind of bookkeeping: every stage is a line with the input, the output and the unit cost at entry sitting side by side. As long as scrap is reduced to a single figure for the month, you can see neither that yields multiply nor that a part at assembly costs three times more than the same part at moulding. Which stage is eating the profit becomes visible when the same data is collected automatically, which is exactly why small and medium business owners use Finmap: to see the financial picture instead of piecing it together.

Read also

• The shop hour in custom furniture: the price of rework
• Printing cost: how to calculate press hour and makeready

Topic foundation: Management Accounting Explained: What It Is and Why Owners Need It