The Prototyping Process and Its Stages
Why prototype, the fidelity ladder, one question per prototype, the cost of change and how many iterations to plan.
Why prototype
Drawings and calculations tell you what you expect; a prototype tells you what happens. Real parts reveal what no model shows: a connector you cannot reach, a part that vibrates, a user who holds the device the wrong way. The purpose is learning, so the best prototype is the cheapest one that produces the answer.
- Failing early is cheap. A problem found with a cardboard mock-up costs an hour; the same problem found in tooling costs weeks and thousands.
- One question per prototype. A prototype that tries to test everything usually proves nothing clearly.
- Build the riskiest part first. If a mechanism or sensor might not work, test that before designing the box around it.
- Show, do not tell. A rough model communicates to a customer, a manager or a supplier faster than a drawing.
Four kinds of prototype
| Type | Answers | Example | Typical effort |
|---|---|---|---|
| Proof of concept | Can the core idea work at all? | A breadboard with a sensor and a motor that shows the control idea works | Hours to days |
| Looks-like | How big, how heavy, how will it feel and look? | A foam or 3D-printed case with no electronics (see Low-Fidelity Prototyping: Mock-ups and Models) | Hours to days |
| Works-like | Does it do the job under realistic conditions? | A rough enclosure with working electronics and real loads (see Rapid Fabrication: Choosing a Method) | Days to weeks |
| Pre-production | Can it be made repeatably, and does it meet requirements? | A small batch built with production parts and processes (see From Prototype to Production) | Weeks to months |
A focused prototype tests one thing (a latch, a grip, a circuit); an integrated one tests how the parts work together. Most projects need both: focused first, integrated later.
A prototyping process
| Step | What to do | Output |
|---|---|---|
| 1. State the question | Write one sentence: “This prototype will show whether ___.” | A clear success test |
| 2. List requirements | Sizes, loads, speeds, environment, cost, safety | A short list with numbers |
| 3. Choose the cheapest method | Compare options (see Rapid Fabrication: Choosing a Method) | A method and a time budget |
| 4. Build | Keep it rough where the question does not need precision | A working or fitting model |
| 5. Test | Run the planned test and measure, not just look (see Testing, Iteration and Feedback) | Numbers and notes |
| 6. Decide | Keep, change or abandon; write down why | The next question |
| 7. Record | Photos, files and a version number for each iteration | A history you can return to |
How much is a prototype worth?
A prototype is worth building when the value of what it teaches exceeds its cost. The value is the reduction in the expected loss from a wrong decision.
Worked example: is a $150 test rig worth it?
- If a design choice goes wrong it will cost about $10 000 (tooling rework and delay). You think there is a 20 % chance of that: expected loss $10 000 × 0.20 = $2 000.
- A test rig that checks the weak point would reduce the chance to 5 %: expected loss $500. The value of the information is $2 000 − $500 = $1 500.
- The rig costs $150, so the net gain is $1 350: build it. A rig that only brought the chance down to 10 % would be worth $1 000: still more than $150.
- The rule: spend at most the value of the information, and prefer the test that reduces the risk the most per dollar.
The cost of change
Worked example: moving a connector
The same change, found at four stages (taking $5 of effort in the model as the base):
| Change | Stage | Cost (rule of ten) |
|---|---|---|
| Move a connector in the CAD model | Sketch / model | $5 |
| Reprint the case with a moved connector | Prototype | $50 |
| Rework a molding tool | Tooling | $500 |
| Recall or retrofit units in the field | Production | $5 000 |
The numbers are illustrative: the point is the ratio. A cheap model that finds the problem a stage earlier saves ten times its cost.
How many iterations?
Each prototype answers some of the open questions and exposes new ones. Suppose a good iteration resolves about 60 % of what is still unknown.
| Iterations done | Unknowns left |
|---|---|
| 0 | 100 % |
| 1 | 40 % |
| 2 | 16 % |
| 3 | 6 % |
| 4 | 3 % |
| 5 | 1 % |
Worked example: iterations to reach 5 % unknown
- Remaining unknowns after n iterations = (1 − 0.60)n. Setting this to 0.05 gives n = ln 0.05 ÷ ln 0.40 = 4 iterations.
- Budget for that many cycles, and for a few more if the product is complex. Three or four cycles are normal for a simple product, five to ten for a new one.
- Build in parallel when each build is cheap: three concepts at $40 each cost $120 at once and finish in one cycle; building one at a time and being wrong twice costs the same money and two more cycles of waiting.
Habits that keep a project moving
Worked example: time-boxing
- Suppose you have 40 hours of build time and expect 5 iterations: each gets 8 hours, including testing. If an iteration is going to overrun, stop and decide: either the question was too big (split it) or the prototype is too good for its purpose.
- Use off-the-shelf parts for everything that is not the point of the experiment: dev boards, standard motors, sensor modules, brackets and fasteners (see Electronics in a Prototype).
- Keep every version. A file naming scheme (project_v03) and a log of what changed will save you when version 5 turns out worse than version 3.
- Get a second opinion early. Someone who has not seen the project will find what you cannot.
- Photograph and measure each build before you change it.
- Write down the decision, and why: the same mistake made twice is expensive.
Safety with prototypes
- A prototype is untested. Do not leave unproven electronics, heaters, batteries or motors running unattended, and add a fuse or current limit.
- Lithium batteries need protection circuits and correct chargers: they can burn violently if abused (see Fire, Fuels and Explosion).
- Mains voltage prototypes need proper insulation and enclosures, and a competent person to check them.
- Tools and processes have hazards of their own: see Workshop Safety.
- Do not give unreviewed prototypes to others in a form that could hurt them (sharp edges, hot parts, pinch points, mains).
Troubleshooting
| Problem | Likely cause | Try |
|---|---|---|
| Too many iterations, no progress | The question is unclear or too broad | Write the question and the success test; split it into smaller tests |
| Prototype too costly or slow | Too much fidelity for the question | Go back a step on the ladder; use a mock-up |
| Prototype works on the bench but not in use | Different conditions: vibration, heat, supply, noise | Test in the real environment early |
| Cannot reproduce a good version | No records of settings or files | Keep versioned files and a log of materials and settings |
| Everyone loves the prototype, the product fails | Tested only with friendly users | Test with people who match the real customers (see Testing, Iteration and Feedback) |
Related
- Prototyping Techniques: the overview of this guide.
- Low-Fidelity Prototyping: Mock-ups and Models: the cheapest prototypes.
- Project Management: Budget, Time and Risk: planning iterations and budget.
- PCB Prototyping: electronics boards.
FAQ
What is a prototype?
A version of an idea built to answer a question: does it fit, work, please users, or get made. The best one is the cheapest that answers the question.
What is the difference between looks-like and works-like?
A looks-like model shows size, shape and feel with no function; a works-like model does the job with rough appearance. Product teams often build both separately.
How much should I spend on a prototype?
No more than the value of what it teaches: the reduction in the expected loss from a wrong decision.
How many prototypes will I need?
Several: three or four for a simple product, five to ten for a new one. Plan the budget and schedule for them.
Is the rule of ten accurate?
It is an approximation: the real ratio varies, but the cost of a change does rise steeply as the design is frozen into tools and stock.
References
- Ulrich, K. T. and Eppinger, S. D., Product Design and Development, McGraw-Hill.
- Pahl, G. and Beitz, W., Engineering Design: A Systematic Approach, Springer.
- Kelley, T., The Art of Innovation, Doubleday (rapid prototyping culture).
Last reviewed: October 2026. Costs, probabilities and ratios are illustrative: use your own.