The Real Cost of a Wrong Number

In an engineered system a single wrong number propagates into every decision built on it; rigor is cheaper than the failure it prevents.

Errors do not stay put

A wrong number in a coupled engineering design does not stay contained. It propagates into every figure derived from it, every decision that trusted those figures, and every document that quoted them. The effort of catching an error early is small compared with the effort of unwinding everything built on it once it has spread -- which is the practical case for rigor, before any moral one.

How a single error spreads

• Downstream calculations inherit the error silently.
• Design choices lock in around the wrong value.
• Documents propagate it until it looks authoritative by repetition.
• A later measurement disagrees, and now the whole chain is suspect.

Why the standards pay for themselves

Verification, error budgets, regression tests, and review all cost time up front. They are worth it because the alternative -- discovering a foundational error late -- costs far more, and in a program aiming at built hardware around 2030 it can cost trust that is hard to rebuild. Catching a wrong Q or a wrong load in review is inexpensive; catching it after a machine is designed around it is not.

This is the unromantic argument for high standards, and it stands even for someone who cares nothing for scientific virtue. Rigor is the cheapest insurance available against the compounding cost of a wrong number. The moral case, made elsewhere in this section, sits on top of a plain engineering one. The compounding is the whole danger: an error caught in review is a corrected line, while the same error caught after a machine is designed around it is a question about everything that touched it.