Two Tools That Answer Different Questions

Teams treat this as a budget choice: simulate because it is cheaper, or build because it is real. It is not a budget choice. Simulation and physical prototypes answer different categories of question, and using one to answer the other's question is how projects get expensive surprises late.

The useful framing is what you are uncertain about. If the uncertainty is physics you can describe with equations and known material properties, simulate. If the uncertainty involves people, manufacturing variation, or a phenomenon nobody has characterized, build.

What Simulation Genuinely Proves

Analysis is excellent at bounded, well-posed physical questions, and it is fast enough to test twenty variants where you could only afford two physical builds.

  • Structural behavior. Stress, deflection, and buckling under a defined load in a metal or a well-characterized plastic. A bracket that must hold 80 lb without deflecting more than 0.04 in is exactly the right question for FEA.
  • Modal and vibration response. Finding a natural frequency that sits inside your excitation band is far cheaper on screen than on a shaker table.
  • Thermal. Steady-state and transient temperatures inside an enclosure, airflow through vents, and heat sink sizing. Thermal simulation correlates well when the heat sources are known.
  • Mold filling. Mold flow analysis predicts weld lines, air traps, sink, and warpage well enough to move a gate before steel is cut, which is worth thousands.
  • Tolerance behavior. Statistical stack-up across an assembly, testing thousands of virtual builds, which no prototype run can match. See tolerance stack-up analysis.

Cost runs $2,500 to $12,000 for a focused study on a single assembly, with results in days rather than the two to four weeks a physical build and test cycle takes. The economics are covered in FEA simulation in product design.

Where Simulation Quietly Misleads

Every analysis is only as good as its inputs, and several inputs are routinely wrong in ways that are invisible in the pretty color plot.

Material data for plastics is the biggest offender. Datasheet values are measured on injection-molded test bars at room temperature and low strain rate. Your part has weld lines, variable fiber orientation, residual stress from molding, and will sit at 100 degrees F in a car. Simulated strength on a glass-filled nylon part can be optimistic by thirty percent or more for exactly these reasons.

Boundary conditions are the second. A model constrained more rigidly than the real assembly reports lower stress and less deflection than reality. Contact friction, bolt preload, and adhesive stiffness are usually assumed rather than measured. And fatigue, which is what actually breaks products in service, depends on surface finish and local geometry at a level of detail most models do not capture; the failure mode is described in material fatigue.

Drop simulation deserves specific caution. It is genuinely useful for comparing two designs against each other, and genuinely unreliable as an absolute predictor, because impact behavior depends on orientation, floor stiffness, and internal component motion. Use it to rank options, then confirm on a real drop tester as described in designing a product to survive a drop test.

What Only a Physical Prototype Can Answer

No simulation tells you whether a grip feels right, whether a button has satisfying travel, whether a lid sounds cheap when it closes, or whether a first-time user pushes the wrong thing. Those are the questions that decide whether a product sells, and they are only answerable by handing an object to someone.

The physical build also catches the compound problems. Assembly order that turns out to be impossible. A cable that has nowhere to route. Tolerance interactions with the actual parts you can actually get. Thermal behavior with the real enclosure sitting on a real desk in a real room, which differs from the simulated version because nobody models a tablecloth blocking the bottom vent.

Certain domains are effectively physical-only. EMC behavior, regulatory testing, sealing performance, chemical compatibility, and long-duration wear all require hardware. And investors, retail buyers, and potential customers make decisions with their hands; a rendering does not close a deal the way a working unit does, which is the whole argument in looks-like vs works-like prototypes.

Cost, Time, and Who Gets Convinced

A basic printed appearance model costs $200 to $900 and arrives in a week. A functional prototype of a moderately complex product costs $8,000 to $40,000 and takes six to twelve weeks. A focused simulation study lands in between on cost and well below on time.

But look at who has to be convinced. Engineers accept analysis. Regulators require test reports. Customers, retail buyers, and most investors need to hold something. If the next milestone is a funding round or a purchase order, simulation does not substitute, however rigorous it is.

The Order That Usually Works Best

Neither tool comes first as a rule. The productive sequence interleaves them, cheapest and most informative first.

Start with a crude physical mock, even foam or cardboard, to settle size, proportion, and basic ergonomics. This costs almost nothing and prevents you from simulating the wrong geometry. Foam models remain the highest return per dollar in early development.

Then simulate the decisions that are expensive to get wrong: wall thickness, rib layout, gate location, heat sink sizing, and any structure that would need a tooling change. Use analysis specifically to reduce the number of physical iterations, not to eliminate them.

Then build the informed prototype, with the geometry that analysis already narrowed down. Test it against the same load cases you simulated, and compare. That comparison is the step teams skip and should not, because it tells you how much to trust the model. When simulated and measured deflection agree within fifteen percent, the next simulation is trustworthy and you can iterate on screen with confidence. When they disagree by a factor of two, you have learned something important about your assumptions before it cost you a mold. Finally, put the hardware in front of real users, following user testing with a prototype.

Deciding What to Simulate and What to Build

Projects House runs both sides of this: structural, thermal, and mold flow analysis where it saves an iteration, and functional prototypes where only hardware will settle the question. Send your design and the specific thing you are unsure about through our contact form and we will tell you which tool answers it.