A 70% scrap rate does not cost you 70% more material
A process that scraps 70% of what it is fed needs 3.33 kg to ship 1 kg — not 1.70. The difference is the single largest term in a machined metal part's footprint, and it is routinely computed the wrong way round.
5 min read · last checked
Ask what a 70% scrap rate costs in material and the intuitive answer is 70% more. It is wrong, and it is wrong by a factor of two at the end of the range where it matters most.
Scrap is quoted on the input— “we scrap 5%” means five of every hundred kilograms that go in. So to ship one kilogram you must feed 1 / (1 − s), not 1 + s.
| Scrap | 1 + s (wrong) | 1/(1 − s) (right) | Understated by |
|---|---|---|---|
| 5% | 1.05 kg | 1.05 kg | 0.3% |
| 20% | 1.20 kg | 1.25 kg | 4% |
| 40% | 1.40 kg | 1.67 kg | 19% |
| 70% | 1.70 kg | 3.33 kg | 96% |
The two agree while scrap is small, which is exactly why the error survives.
The wrong form is close enough to be believed at low scrap and out by nearly half at high scrap. That is the worst possible shape for an error: it is never caught on the cases where it does not matter, and it is never questioned on the cases where it does.
Why this is the largest term, not a correction to it
Machining carries a 15–70% scrap band. Take the top of it and a 1 kg aluminium part is cut from 3.33 kg of billet — 2.33 kg becomes swarf. Primary aluminium is around 9 kg CO₂e/kg depending on which library you ask (the comparison is here), so:
| Term | kg CO₂e | Where it comes from |
|---|---|---|
| ~21 | metal that becomes swarf | 2.33 kg × ~9 kg CO₂e/kg |
| ~0.22 | the machining electricity | 1.65 kWh/kg × 0.131 kg CO₂e/kWh |
Two orders of magnitude apart. Only the smaller one appears on an electricity bill.
The material scrapped at any step carried its entire cradle-to-gate burden before it was discarded. It was mined, smelted, cast and shipped exactly like the material that ships. A model that counts only the energy a process draws misses the term that dominates it — and it misses it precisely on the operations where the client most needs to know.
Yields multiply along a route. They do not add.
A second, compounding error. Take a real three-step route — mould, machine, coat — and sum each step’s overhead:
| Step | Scrap | Its own 1/(1−s) − 1 |
|---|---|---|
| Injection moulding | 5% | 0.053 |
| Machining | 30% | 0.429 |
| Coating | 20% | 0.250 |
| summed | — | 0.731 kg |
| actually | — | 0.880 kg |
Cumulative yield is 0.95 × 0.70 × 0.80 = 0.532, so feed is 1/0.532 = 1.88 kg.
The summed form understates by 20%, and it understates the energy by more. A step near the front of a route has to process everything that every later step will eventually throw away, so the first operation runs on 1.88 kg to ship 1 kg, not on 1.05.
One consequence is worth stating plainly: reversing two steps in a route leaves the material bought identical — multiplication commutes — and changes the electricity, because a different operation is now the one processing the most. Route order is not cosmetic.
Regrind reduces the purchase, not the throughput
Where scrap is recoverable in house — most polymer processing, most metal — recovering it shrinks the purchase order. It does not shrink what the machines saw. The material was still moulded, still machined, still heated, and the energy for all of that was still spent. Crediting energy for regrind is how a recycling claim ends up paying for itself twice.
Each of the 101 process models states whether its scrap is normally regrindable for this reason, and the figure for material bought per kilogram shipped is given without assuming any recovery — because how much a particular plant actually recovers is a fact about that plant, not about the operation.
What to do with this
Two things. Compute the feed as 1/(1−s) and compound it along the route. Then notice which lever is actually large: on a machined part, a five-point improvement in yield is worth more carbon than any plausible change to the machine, the tariff or the tooling — and near-net-shape upstream beats all of them.
Check it yourself
Every figure above is on this site with its source, geography, method and licence.
Also worth knowing
- Why two published carbon factors for the same material disagree
- The energy factor you quoted is missing about a third of itself
- Your renewable tariff is zero and 0.131 at the same time
- The published factors say gas is 15% worse. Delivered as heat it is 2–3.5×.
- The same steel is 1.99 and 1.14 kg CO₂e, and both are correct
- Some of a kiln's CO₂ comes out of the rock, not the flame
- When cutting carbon makes something else worse
Use this
These are the mistakes the assessment engine is built to avoid making. To watch it not make them on a product of your own, run a bill of materials through the free carbon calculator — no sign-up, and it runs in your browser. For the whole picture, including the process and its scrap, model the whole route in the prototype, or see what TrueGauge does for a manufacturer.
Early access
Request early access.
Both products are pre-launch. We're working with an early group of manufacturers, architects, builders and sustainability teams to shape what gets built first — tell us which platform is relevant to you.
Shape the roadmap
Tell us which decisions you need supported, and what the data has to do to be usable.
First access
Be first in line as each platform opens to early users.
A direct line to the team
Early conversations happen with the people building it, not a sales funnel.