Nominal stress is a comfortable fiction: load over area, spread evenly. Real parts have holes, grooves, keyways and shoulders, and at every one of them the stress lines crowd together and the local stress leaps. The bookkeeping is deliberately simple: $\sigma_{max} = K_t \, \sigma_{nom}$ — sigma-max equals K-t sigma-nom. $\sigma_{max}$ is the peak elastic stress at the notch root in MPa; $\sigma_{nom}$ is the nominal stress on the NET section — the reduced section through the hole, not the full width; and $K_t$ is the stress concentration factor, a naked number with no units, read off a Peterson chart for that geometry.

The famous value is 3: a small circular hole in a wide plate under tension triples the stress at the sides of the hole. Generous fillets sit near 1.5, sharp ones near 3, and a square inside corner runs away to numbers you should not design to at all. Kt depends only on GEOMETRY — not on the material, not on the load. It is the shape's own multiplier.

Two honest footnotes. Kt is an elastic factor: a ductile metal that yields locally at the notch root redistributes the stress and often survives happily, which is why static design frequently ignores Kt for ductile parts. Fatigue never ignores it — the great majority of fatigue cracks start at a stress raiser, and the alternating stress that feeds a fatigue check is the notch-amplified one. Multiply first; compare second.