Price per kilogram is the worst polymer procurement metric

Three quotations sit in the tender table, sorted by currency per kilogram. The cheapest is highlighted in green. The trouble is that your shop does not sell kilograms — it sells parts. And between a kilogram of pellets and a part in a box stand four multipliers, none of which made it into that table. This article shows how to calculate cost per part instead of price per kilogram, and why a "per kg" specification structurally selects the worse supplier.

Every numeric range below is given either with a reference to a standard or with the note "typical for the class — verify against the grade TDS". The arithmetic examples are built on explicitly declared inputs: substitute your own figures — the method will not change. The aim of this article is not to hand out universal coefficients but to give the reader a working model.

You buy kilograms and sell cubic centimetres

This is not a rhetorical flourish but literal accounting. The drawing fixes the volume of the part — it is set by geometry and does not change with what you mould it from. Part mass equals that volume multiplied by material density. Therefore the number of parts from a tonne of pellets is determined by density, not by price.

In materials science this is well established. In his classic material selection methodology Ashby states directly: to move from a minimum-mass index to a minimum-cost index, density ρ in the formula is replaced by the product C_m·ρ — the cost per unit volume (Material selection, methodology overview; Ansys Granta, on Ashby plots). Price per kilogram in its pure form appears in no correct selection index — it always travels paired with density.

Density is determined per ISO 1183 and, for filled grades, follows the rule of mixtures predictably:

Substitute ρ(PA6) ≈ 1.13 g/cm³ and ρ(E-glass) ≈ 2.54 g/cm³. For GF30 we get ≈ 1.36 g/cm³, for GF50 — ≈ 1.56 g/cm³. Both values fall inside the ranges we hold as class control values (PA6 GF30: 1.35–1.40 g/cm³), so the model works. Here is what that means in cubic centimetres per kilogram.

Read it like this. If supplier A offers unfilled PA6 and supplier B offers PA6 GF50 at the same price per kilogram, then a part made from material B costs you 1.38 times more in raw material (1.56 / 1.13). To break even on cost per part, GF50 would have to cost about 0.72 of the unfilled PA6 price per kilogram. That practically never happens on the market — but it does not follow that GF50 loses. The comparison has simply not started yet: we have counted only the first of four multipliers.

Four multipliers between pellet and part

where ρ is density, V is part volume per the drawing, C_kg is the pellet price per kilogram, b is the scrap fraction, t_cycle is cycle time, S_machine is the machine-hour rate and n is the number of cavities.

The formula is deliberately simple: there is nothing in it you do not already know. Its whole value is that price per kilogram sits in it as one factor out of five, and it controls none of the other four.

Part volume is not a constant if you change the material

Here hides the most expensive mistake, and because of it the "per kilogram" comparison is not merely imprecise but systematically biased.

Part volume is fixed only as long as you do not touch the drawing. But a part is usually designed for stiffness. The bending stiffness of a plate or a rib is proportional to the product E · s³, where E is the modulus and s is the wall thickness. If you replace a stiffer grade with a cheaper, less stiff one, the only way to keep the assembly stiffness is to thicken the wall — and thickening enters as a cube.

Take flexural modulus values typical for the class: PA6 GF50 — about 16 GPa, PA6 GF30 — about 9 GPa (both typical for the class, verify against the TDS). The thickness ratio for equal stiffness:

The wall thickens by 21 %. Part mass then grows as ρ · s, that is (1.36 · 1.21) / (1.56 · 1.00) ≈ 1.06 — by about 6 %. So the "saving per kilogram" from switching to GF30 is partly eaten up at the mass stage already. But the main point is ahead: cooling time grows as the square of wall thickness, that is by roughly 47 %.

Note the asymmetry: stiffness rewards us as a cube, while cooling punishes as a square. This is exactly why the design choice "take a stiffer material and make the wall thinner" so often wins on cost, despite the higher price per kilogram.

Scrap — the multiplier nobody writes into a quotation

Formally, scrap multiplies material cost by 1/(1 − b). The difference between 2 % and 6 % scrap means coefficients of 1.020 and 1.064, that is about 4 % difference in material consumption. At this point the buyer usually waves it off: "four percent is trivial against a ten percent discount per kilogram."

The mistake is that a scrapped part takes more than the pellet. It takes a full machine cycle, a cavity slot, operator labour and, on critical parts, inspection time as well. The real price of a scrap percentage is therefore tied not to material cost but to the machine-hour rate divided by the number of good parts. And the more expensive the machine and the shorter the cycle, the more expensive every percent of scrap becomes.

Here is how the four multipliers behave on a material change — and by which method each of them is verified.

Why a "per kilogram" tender structurally selects the worse supplier

This is the most uncomfortable part, and it is not about bad faith but about the geometry of the criterion.

Picture two suppliers. The first holds a stable MFR between batches, issues a CoA with every batch, selects the grade for your wall thickness and supports the mould trial. All of that costs money and sits in the price per kilogram. The second offers the nominally identical grade 8 % cheaper — through a wider viscosity tolerance, a less robust stabiliser system or a higher share of reprocessed material in the compound.

In the "per kg" column the second one wins. In the "cost per good part" column he may lose — through a longer cycle, a wider dimensional spread, an extra percent of scrap on weld lines. But that column does not exist in the tender, and only the customer can create it.

Hence the practical conclusion: the criterion "lowest price per kilogram" is not a neutral way to compare but an active filter that screens out the suppliers whose value lies in the process multipliers. If you want such suppliers to come to your tender, the criterion must contain at least density, and better still cost per good part by an agreed method.

The cheapest way to fix a tender: add two mandatory fields to the quotation form — density per ISO 1183 and MFR per ISO 1133 with the test conditions stated. It costs nothing, discloses nothing and already makes the comparison fair to a first approximation. How to read the rest of the data sheet we covered separately: How to read a polymer TDS.

When price per kilogram is relevant after all

It would be dishonest to push the thesis to absurdity. There are situations in which price per kilogram is a perfectly workable metric.

• The product is sold by weight: pipe, profile, sheet, cable sheathing. Here mass is the product.
• The material goes into compounding or recycling rather than into a part with a drawing.
• Two batches of the same grade from different suppliers are compared — density, modulus and rheology match by definition.
• The part is thick-walled and slow, the machine-hour is cheap, and material is the dominant share of cost.

• Comparing grades with different filler type or content by price per kilogram makes no engineering sense — it is a comparison of different volumes.
• Comparing different polymers (PA6 versus POK, PA6 GF30 versus PA66 GF30) by price per kilogram makes even less sense: density, shrinkage, modulus and thermal diffusivity all differ at once.
• A thin-walled fast-cycling part on an expensive machine is a case where material price may turn out to be a secondary term. But that is precisely the case that must be calculated, not assumed.

Polyketone is a telling example of how the arithmetic works. Exablend® POK has a density of about 1.24 g/cm³ (ISO 1183) — between unfilled polyamide and PA6 GF30, so it yields more parts per tonne than a glass fiber reinforced grade. At the same time its shrinkage is markedly higher (roughly 1.5 % longitudinal / 1.6 % transverse per ISO 294-4), and that is a requirement to recalculate tolerances rather than a "free" advantage. We covered separately what exactly the designer has to recalculate: Designer's guide: POK instead of POM and polyamide. The cost of a POK part may turn out lower or higher than a polyamide one — and that is always a question for the specific drawing, not for the price list.

What to check before a production run

• Density per ISO 1183 in the TDS of every candidate grade, not "I read somewhere that it's about the same".
• Flexural modulus and wall thickness recalculated for equal assembly stiffness, not carried over from the old drawing.
• Longitudinal and transverse shrinkage per ISO 294-4 — and whether the part meets tolerance without mould rework.
• Actual cycle time on a trial batch, not a catalogue estimate; separately, the cooling time of the thickest section.
• Scrap rate measured on a batch of sufficient size, broken down by cause (warpage, weld line, short shot).
• MFR spread between batches per the CoA — a source of hidden scrap in series production.
• Pellet moisture before moulding: for polyamides this is a precondition for the reproducibility of every point above.
• Abrasive wear of the screw and hot runner when switching to a highly filled grade — an expense that surfaces months later.

Expert review: 5 questions about cost per part

1. Why can't a tender simply normalise price by density and stop there? Because normalising by density closes the first of four multipliers and says nothing about the rest. It will correctly compare two grades with the same modulus and the same rheology — two batches of PA6 GF30 from different suppliers, for instance. But as soon as the candidates differ in stiffness, you are not entitled to hold wall thickness constant, and part volume therefore stops being a common basis for comparison. Normalising by density is a correct first step and a poor last one.

2. How much does a filled grade actually shorten the cycle? The direction is stable: as glass content rises, thermal diffusivity usually rises, and higher heat resistance permits a higher ejection temperature — both effects shorten cooling. The magnitude cannot be named: it depends on the grade, the thickness, the cooling system and the acceptance criterion for the part at ejection. Anyone quoting you a specific percentage without your mould is selling, not calculating. The correct action is two trial batches and a stopwatch.

3. We mould a thin-walled part on an expensive machine. Is it worth haggling over pellets at all? It is, but with a different priority. In that regime the machine-hour dominates, so the biggest money lies in cycle stability and in the scrap rate, not in the price per kilogram. The rational strategy is to pay for predictable rheology (a narrow MFR tolerance, a CoA per batch) and for a high-flow grade that lets you lower pressure and melt temperature. A discount per kilogram bought at the price of drifting viscosity is almost always a loss here.

4. Do runners and reprocessed material change the picture? Yes, and in both directions. Returning regrind reduces effective virgin pellet consumption, but every reprocessing cycle shortens the molecular chain and, for glass fiber reinforced grades, shortens the fiber length — and modulus falls with it. A falling modulus takes us back to the wall thickness point: to hold stiffness you will have to either limit the regrind share or thicken the wall. The regrind share is therefore determined by testing the part, not by a wish to save on procurement.

5. What if a supplier refuses to give density and MFR with the test conditions? Treat that as a substantive answer. Density per ISO 1183 and MFR per ISO 1133 with the temperature and load stated are baseline rows of any technical data sheet; they are not a trade secret and do not disclose the formulation. A supplier who does not present them in standardised form either has no batch control of his own or is not prepared to take responsibility for reproducibility. Both are pricing information — just not in the column where people are used to looking for it.

Summary

Price per kilogram is not a wrong number. It simply answers a question your production does not ask. Production asks what a good part in a box costs, and that question is answered by four multipliers: density, part volume (which depends on modulus), cycle time (which depends on the square of wall thickness) and scrap.

The good news is that all four come from open sources: the TDS, the drawing, a stopwatch and the scrap log. The bad news is that none of them will appear in the comparison table by itself. It has to be put there.

Examid® PA6 GF50 R10 — Maximum reinforcement — a tool for cutting wall thicknessModulus ≈ 16 GPa · density typically ≈ 1.56 g/cm³ (ISO 1183) · TDS on requestRequest supply terms →

Examid® PA6 GF30 — The baseline series moulding grade — a control point for calculationFlexural modulus 8–10 GPa · density 1.36–1.38 g/cm³ · CoA per batchRequest supply terms →

Exablend® POK M330A — Polyketone at ≈ 1.24 g/cm³ — more parts per tonneISO 1183 · shrinkage per ISO 294-4 requires tolerance recalculationRequest supply terms →

Material Wizard supplies Examid® engineering polyamides and Exablend® polyketones, provides a TDS and a CoA per batch and supports technical grade selection for a specific part — including a cost-per-good-part calculation against your drawing and your machine-hour rate. The company is located in Derazhnia and Kharkiv. The material is available with delivery across Ukraine — to calculate the cost per good part for your specification, please consult our specialist.