Prototyping and iteration

Why iteration beats getting it right first time

Most design faults are not visible on a screen. A bracket that is clearly stiff enough in a model rings under vibration. A housing that assembles perfectly in CAD cannot be assembled by a person with two hands. A cover fits and then fouls the thing next to it. A spanner will not reach the rear bolt.

These are found by holding the part, offering it up, and seeing what is wrong with it — and each of them costs almost nothing to fix at prototype stage and a great deal to fix later. The point of a printed prototype is not that it is a finished part. It is that being wrong stops being expensive.

The industrial mount job on this site is exactly this: many versions printed, more expected, a design being arrived at rather than handed over. It is the least photogenic part in the gallery and the clearest example of how the work runs.

What a round looks like

Start from wherever you are. A finished STL, a solid model, a dimensioned drawing, a sketch, or a conversation about what the part has to do. None of these is the wrong starting point, and no file is required to begin.

Print, and print what matters. A prototype does not have to be the whole thing. Where only one interface is in question, printing that section is faster and cheaper and answers the question just as well.

Hold it and be honest about it. The most useful feedback is specific: this is too tight, this flexes, this is unreachable, this is the wrong way round.

Revise from the last version, not from scratch. The model is kept and versioned between rounds, so revision five starts where revision four finished. What changed between versions is recorded, because by the ninth nobody reliably remembers what the fourth was for.

Keep the printing constant. Same material and same settings between versions unless there is a reason to change one, so that when a prototype behaves differently the design is the reason and not the printing.

Checking a design before it is committed

Two things beyond printing are available, and both are described here exactly as far as they go.

Simulation can be run where it helps — to compare two ways of doing something, or to see whether an approach is sound before committing to it. It is an indication and nothing more. A printed part is layered, its strength depends on orientation and infill and on how well the layers bonded on the day, and no analysis of a solid model captures that. It is useful for direction, and it is not a substitute for testing the real part. It will never be presented here as though it were.

Destructive testing can be arranged for smaller parts — loading and impact, at small scale. It is a controlled and genuinely informative check for the kind of duty these parts see. It is not certification, it is not testing to a published standard, and it is not aerospace-grade validation. This is not an aerospace workshop. Where a part matters that much, it needs a testing house and a different manufacturing process, and you should hear that rather than be reassured.

The honest limits

A prototype is not a production part. Printed parts behave differently from moulded or machined ones — different stiffness, different failure modes, different surface. A printed prototype answers questions about fit, ergonomics, assembly, clearance and general behaviour extremely well. It answers questions about the exact strength of an injection-moulded final part poorly.

Materials are PLA, PETG and flexible TPU. If the production part will be a different material, the prototype tells you about geometry rather than about material performance, and it is worth being clear which question you are asking.

Your design stays yours. Anything you supply is used for your job only and never reused. Where we model for you, the model comes with you — see the file policy.

Who this suits

Businesses developing a product or a piece of equipment, engineers working through a design, and anyone whose next decision depends on holding the thing. Funded development with several rounds ahead of it is the best fit, and it is welcome as ongoing work rather than a series of unrelated orders.

Minimum order is $75. Prototyping is cheap per round, not free per round, and knowing that up front saves everybody a conversation.

Frequently asked questions

How many rounds does a design usually take?

There is no useful average — it depends on how much is unknown at the start. What matters is that each round is cheap and fast enough that the number stops being the thing you are optimising. A design that takes six rounds and ends up right is a better outcome than one that takes two and does not.

Can you print just part of my design?

Yes, and it is often the right call. If only one interface, one bracket or one section is in question, printing that alone is faster and cheaper and answers the same question.

Do you keep my model between revisions?

Yes. It is kept and versioned so a revision starts from the last one rather than being rebuilt, and what changed between versions is recorded. Your files are never reused for anyone else.

Can you tell me if my design will be strong enough?

Partly, and only with caveats. Simulation can indicate whether an approach is sound, and destructive testing at small scale can tell you something real about how a printed version behaves. Neither is certification, and neither predicts the behaviour of a part made by a different process in a different material. You will get the honest version of what the check does and does not prove.

What if I only have an idea, not a file?

That is a normal starting point. One of the jobs on this site began with nothing but a conversation — see the pump housing case study.