Iterative design and prototyping means building the cheapest physical version that answers one question, testing it, feeding the result back into the design, and repeating until the open questions are small enough for production to settle. It works because late changes are the expensive ones, costing more and delaying more than early ones, and a physical part exposes problems a CAD model hides.
The loop only pays off when each round is small and pointed. Stanford Biodesign’s prototyping guide calls the most common novice mistake building the whole concept from the start instead of building up from single parts. Each round should go after the biggest technical risk still open.
How Is Iterative Design and Prototyping Different From a Linear Process?
A linear process finalizes the design, builds a prototype, tests it, and starts over if it fails. The first physical test comes last, so bad geometry, parts that will not assemble, and a grip that feels wrong all show up when fixing them means reopening decisions everything else was built on.
An iterative process moves the first physical test to the start. The Stanford guide says early prototypes should be simple, quick, cheap, and incomplete, with more time and money added as each round retires a risk. Teams that build only what answers one or two narrow questions at a time make quicker progress than teams that build everything at once.
What Should Each Round of Prototyping Test?
The Stanford guide sorts prototypes by what they prove, from least to most burdensome: looks-like, feels-like, works-like, and is-like. A looks-like model in foam or printed plastic answers questions about size, grip, and how parts relate, and it is the quickest to make. Some looks-like models are good enough to hand to users for feedback on sizing and layout. A works-like prototype only has to work well enough to answer one question, and most of them barely work.
The is-like prototype, close enough to the final product that someone could actually use it, comes last. The guide notes that it is typically not the right time to build one until well after you think it is. We would write the question for the round down before anything is printed or cut, so the build gets judged on the answer instead of on how it looks. And we would start with whichever assumption would hurt most if it were wrong, since that is the risk worth retiring first.
A rough concept has a long way to go before it is a manufacturable part. Our napkin sketch to manufacturable part walkthrough follows that path, and getting started on prototyping covers how to choose the first build.
Which Process and Material Fit Each Stage?
3D printing suits the early rounds. It needs no molds or fixtures, and for simple parts in small numbers FDM (fused deposition modeling) is the cheapest way in, so a wrong turn costs a print instead of a tool. Machining costs more for a one-off because its programming and setup land on a single part, which is why the switch to production material deserves a deliberate plan.
The material becomes the question once a test is about strength or heat. FDM parts are weaker between layers than within them, often by 30 to 50% according to Formlabs, and printed PLA softens around 55 to 60°C. A printed part’s result on a load or heat test does not carry over to a part made from the production material.
We would switch to the production material as soon as the question is about strength, heat, or wear. Kemperle prints in FDM, SLA, and SLS and cuts parts by CNC routing, so early and late rounds can be built without changing vendors. Our guide to how rapid prototyping works compares the tolerances and turnaround of each method.
How Should the CAD File Change Between Rounds?
Iterative design does not mean abandoning CAD. The model evolves alongside the prototypes, and every physical test should produce a specific, recorded change to it. Say a bracket cracks in round two: the useful record reads “wall thickened from 2 mm to 3.5 mm in the load-bearing section, after the round-two load test,” where “made it stronger” tells the next round nothing.
The Stanford guide gives the reason for keeping that record: without a history of each iteration, it is hard to tell which features improved performance and which degraded it. A dated copy of the CAD file saved before each round is enough, and for a mechanical build a few annotated photos, a short caption, and a parts and materials list complete the record. When a later prototype performs worse than an earlier one, that history is how you find the change responsible. It also helps when the people who model the part also make the prototypes, because the person who sees the crack is the person who edits the file.
When Is It Time to Stop Iterating?
Stop when the prototype passes its functional requirements every time and the remaining open questions are ones early production can answer better than another prototype. That means at least one prototype close to production in material, process, and geometry. The Stanford guide calls that is-like prototype the flag on a hill: the destination a team aims for, with several other hills to climb on the way.
The failure on the other side is scope creep. The Stanford guide describes a balance between over-building, meaning materials or features the question does not need, and under-building, meaning a prototype that cannot yield a useful result. A test plan that traces each test to the requirement it checks keeps a team on the right side of that line.
Here is the gut check we would use. If the last round’s changes came from real test failures, keep going, and if they came from someone’s opinion of how the part could look or feel with no test behind it, the loop has stopped paying for itself. The common rapid prototyping mistakes are worth a read before your first round.
Send the current CAD file and the question your next round has to answer through the contact page. If you would rather talk it through first, the team is at 718-557-9578.
