Pull a 6 mm steel rod to destruction in a tensile machine and the stress–strain curve hands you two numbers people love to confuse. The bar yielded at about 389 MPa — the point where it stops springing back — and it didn’t actually fracture until about 571 MPa. A roughly 47% gap between the stress it survives and the stress where it stops behaving. The tempting headline is the bigger number. You design to the smaller one.
Here’s why the gap is a trap rather than a bonus. Below yield, steel is elastic: load it, it stretches; unload it, it returns. Above yield it’s plastic — every bit of extra stress buys permanent deformation you can’t take back. The bar happily climbs all the way to 571 MPa, but everything past 389 is the metal slowly failing while still holding load. The strain numbers make it vivid: strain at yield was about 0.0036, strain at fracture about 0.24 — the rod stretched on the order of sixty times more getting from yield to break than it did reaching yield in the first place. That huge “extra strength” is the material coming apart in slow motion.
The aluminium specimen I tested alongside it told the same story — elastic, then a long plastic march to ~16% elongation before it let go. Different metal, same shape of curve, same moral: the ultimate tensile strength is what a coupon survives once, on the way to the scrap bin. It is not a property you get to use.
So the lesson generalizes well past metal: the failure load is not the design load. The usable limit of almost anything — a beam, a bolt, a bracket, even a schedule or a budget — is the point where its behaviour turns irreversible, not the point where it finally breaks. Past yield the part hasn’t failed yet, but it’s no longer the part you designed; it’s a bent thing that happens to still be in one piece. Keep your margins below the yield, and the worst case is a part that springs back.