A Design Is Not Valid Until It Is Tested
January 14, 2026
On our monocoque project, the car body looked right from every angle. The proportions were good, the client liked the direction, and the surfaces were clean. Then we ran the CFD.
The headlight area was costing us about 15% in drag. Nothing about it looked wrong. It was a small change in surface shape that caused the airflow to separate. You could stare at that model for a month and never see it.
Simulation and physical models answer different questions
CFD told us how air would move over the body. It did not tell us how the car would feel standing next to it. For that we printed small scale models and put them on the table. Proportions that look fine on a monitor can look off in your hands, and the reverse is true too.
We ran both in parallel. When the simulation and the physical model agreed, we moved on. When they did not, we stopped and figured out why.
What we learned
- Test early, when changes are cheap. Fixing the headlight surface in CAD took an afternoon. Fixing it after the molds were made would not have been an option.
- Use the right test for the question. Simulation is great at physics. Physical models are great at proportion and fit. Do not ask one to do the other's job.
- Trust the data over your eye. The version we liked most was not the most efficient one, so we changed it.
- Know what your model leaves out. Every simulation has assumptions. Write them down so you know what has not actually been tested.
Where it ended up
Once the design was validated, manufacturing was straightforward. Full size panels were 3D printed and used as masters for the molds, and the final shell was laid up in fiberglass and carbon fiber. The body came in around 40 kg and fit the client's existing mechanical layout.
The lesson we keep relearning is simple. Looking right and being right are different things. Test before you commit.