Vacuum Casting at Small Batch: Running 20, 50 or 200 Parts
Between "one prototype" and "open a mould" there is a real production zone. It has its own cost structure, its own schedule logic, and its own failure modes - most of which are labour, not tooling.
What "small batch" actually means
In vacuum casting, small batch is roughly 10 to 200 pieces. Below ten, you are still prototyping and the question is whether the design works. Above a few hundred, injection moulding starts to win and vacuum casting becomes the expensive way to avoid a tool.
The awkward part is in the middle. At 50 pieces you have a real product to put in front of customers, a real schedule, and a real consistency problem - but you do not have a steel tool, and the process you are using is closer to craft than to manufacturing.
That gap is where small-batch projects fail: not because the casting is bad, but because nobody planned it as a production run.

The cost step nobody expects
A silicone mould does not last forever. Depending on geometry and resin, one mould yields roughly 20 parts, and the practical range runs from 10 for thin, complex parts to 50 for simple blocks.
That single number sets the shape of your whole project. For 200 parts you do not need a bigger mould. You need ten moulds, all cast from the same master pattern.
|
Quantity |
Moulds needed |
What changes |
|
1–20 |
1 |
Prototype territory. Judge the design, not the process. |
|
20–50 |
2–3 |
Field trials. Consistency starts to matter. |
|
50–100 |
3–5 |
Pilot production. You need a schedule and a record. |
|
100–200 |
5–10 |
Small-batch manufacturing. Plan mould rotation and resin lots. |
|
200+ |
- |
Compare injection moulding seriously before proceeding. |
The important consequence: the per-part price does not fall smoothly. It drops when a mould is paid off, then jumps when you need the next one. If a supplier quotes 200 pieces at the same unit price as 40, either the mould life is much longer than claimed, or the price is a guess.

Schedule: the constraint is the operator, not the machine
Casting is a manual cycle. Mix, degas, pour, cure, demould, trim. A polyurethane system typically needs two to four hours in the mould before it can be handled, so a single mould gives you two or three cycles a day - not fifty.
A worked example, with assumptions stated:
20 parts per mould
2 cycles per day per mould
one operator running four moulds in rotation
That gives about 8 parts a day, or roughly 25 working days for 200 pieces. Add the master pattern and the first mould and you are at five to six weeks.
Change one assumption and the answer changes. Two operators and ten moulds in rotation can cut that substantially, because the curing time overlaps. But the curing time itself cannot be compressed by working harder - only by choosing a different resin system, which changes the part's properties.
This is why the honest question at the quotation stage is not "what is the unit price" but "what is the delivery date, and how many moulds are you committing to".

Where the money actually goes
|
Cost element |
Behaviour at small batch |
What to watch |
|
Master pattern |
One-off, fixed |
Must survive repeated mould-making; a soft master limits you later |
|
Silicone moulds |
Per 20 parts, step cost |
Mould count dominates the total |
|
Resin |
Per part, plus waste |
Pot life means every mix has scrap |
|
Labour |
Scales with parts, not with moulds |
The real driver at 100+ pieces |
|
Inserts and second operations |
Per part, often manual |
Frequently forgotten in the quote |
|
Inspection |
Sampling, not 100% |
Needs to be defined, not assumed |
At 20 pieces, the master and the first mould dominate. At 200, labour and resin dominate, and the mould cost has become a minor line. That shift is worth knowing before you negotiate.

Consistency: the risk that small batch hides
A prototype can be one part. A small batch is a population, and populations get compared - by your customer, by quality, by the field.
Four things drift in a vacuum casting run:
Resin lot. Two batches of the same grade can differ slightly in colour and hardness. Use one lot for the whole run where the quantity allows, and record the lot number on the parts.
Mix ratio and degassing. Polyurethane is mixed by weight, not by eye, and degassed before pouring. A rushed mix shows up as bubbles at the surface or a soft patch inside.
Mould age. Part twenty from a mould is not identical to part one. The silicone has taken a compression set, so dimensions creep. Rotate moulds deliberately rather than exhausting one at a time, or you will get a batch trend you cannot explain.
Colour. Matching under one light and shipping under another is a classic dispute. Agree the light source and keep a physical reference sample.
The practical answer is not more inspection, it is traceability: mould ID and resin lot on every part, one first-article from each new mould, and a hardness check per batch.

The advantage that justifies the process
Here is what vacuum casting does better than any alternative at this quantity: you can change the design in a day.
The master pattern is the single source of geometry. Revise it, and every subsequent mould follows. Ten moulds do not need ten revisions. In a product development programme where the housing changes three times before launch, that is worth more than the unit price difference against injection moulding.
The limit is the master's own durability. A machined aluminium or a good SLA master will survive many mould pulls; a soft or rough master will not, and the wear shows up as loss of sharp detail and drifting dimensions - usually around mould five, which is exactly when nobody is looking.

Inserts and second operations
Small batches often need features that casting alone cannot give:
Threaded inserts - installed hot or moulded in. Moulded-in inserts need a locating pin in the mould, which must be designed before the master is made.
Bonded joints - an insert glued in after casting is a process step, not a design feature; it needs a fixture and a cure time.
Machined interfaces - a bearing bore or a sealing face is best cast slightly oversize and then machined. Plan the allowance.
Painting and laser marking - text applied after casting will not survive the same wear as moulded-in text, but it is the only option mid-run.
Each of these adds a step that is easy to omit from a quote and hard to add later.

Design rules for a small-batch casting
|
Feature |
Rule |
Reason |
|
Wall thickness |
1–3 mm, kept uniform |
Thick sections shrink and sink; thin ones fail to fill |
|
Draft angle |
1–2° minimum |
Silicone is flexible, but not infinitely |
|
Undercuts |
Allowed with a split mould |
Free complexity, but more labour per part |
|
Sharp corners |
Radius them |
Air traps and stress risers |
|
Deep narrow cores |
Avoid below about 2:1 |
The core bends and the wall goes off-centre |
|
Gate and vent position |
Decide with the moulder |
Determines where the witness mark sits |
|
Shrinkage |
Small for PU, but confirm per resin |
Master is usually cut to final size |
|
Text and logos |
0.3 mm depth minimum, raised preferred |
Engraved detail fills with resin |

When to stop vacuum casting
|
Vacuum casting |
Injection moulding |
CNC |
|
|
Best quantity |
10–200 |
500+ |
1–100 |
|
Tooling cost |
Low, per 20 parts |
High, one-off |
None |
|
Unit cost at 200 |
Moderate |
Lower once tooling is amortised |
Higher for complex shapes |
|
Tolerance |
±0.1–0.2 mm typical |
±0.05 mm typical |
±0.01 mm on critical features |
|
Material choice |
Polyurethane only |
Full thermoplastics range |
Any machinable plastic or metal |
|
Design change |
Days |
Weeks, and tooling cost |
Immediate |
Two honest triggers to leave vacuum casting:
1. The break-even has passed. Once the amortised injection moulding unit price falls below the casting unit price plus its labour, continuing is habit, not engineering.
2. The material is wrong. Vacuum casting gives you polyurethane. If the part needs ABS, PC, PA or PEEK behaviour - impact strength, UV stability, a specific dielectric - no amount of process skill will produce it.

A checklist for a 200-piece run
Freeze the master revision and record it.
Agree mould count and delivery date in writing, not just unit price.
One resin lot for the run, lot number recorded.
First article from each mould, measured on the critical features.
Hardness check per batch, colour against an agreed reference under an agreed light.
Mould ID marked on every part, so a field failure can be traced.
Agreed sampling plan instead of 100% inspection, and agreed acceptance limits.
Packaging that survives shipping without the parts rubbing - cast PU scuffs easily.
About the manufacturer
Shenzhen Strongd Model Technology Co., Ltd. provides vacuum casting, CNC machining and industrial 3D printing for robotics, automotive, medical and consumer electronics customers. The Shenzhen plant covers 40,000 m² and runs 85 CNC machining centres, 18 of them five-axis, with Hexagon CMM inspection under ISO 9001:2015 and IATF 16949:2016. Prototypes ship in 3–7 days; quotations are returned within 8 working hours.
Website: https://www.strongdcnc.com/
