A fillet weld's cross-section is a right-angled triangle with both legs lying on the plates. It does not fail along a leg. It fails across the throat — the shortest way through, from the root to the face — and that is the altitude of the triangle.

$a = 0.707\,z$, read aloud a equals nought-point-seven-nought-seven z. $a$ is the effective throat in millimetres, and $z$ is the fillet leg size in millimetres — the number on the drawing, measured along the plate. The 0.707 is not a fudge: it is $1/\sqrt{2}$, the altitude of an equal-leg right triangle, and codes print the rounded figure so hand arithmetic and software agree to the last digit.

Use the leg where the throat belongs and the weld looks 41 per cent stronger than it is. That single slip is the commonest arithmetic error in weld design, and it always errs generously, which is why it survives review.

Capacity then follows in one line: $F = \tau A_t$ — F equals tau A-t. $F$ is the load the weld can carry, $\tau$ — the Greek letter tau — is the allowable shear stress on the throat in megapascals, and $A_t$ is the throat area in square millimetres, the throat thickness times the effective length. A fillet is checked in SHEAR on its throat whichever direction the load pulls; that convention is shared by every structural code, and it is why one number covers every case.

Two honesties. A megapascal is a newton per square millimetre, so MPa times mm² is newtons with nothing to convert — that is why weld work is done in millimetres. And the allowable belongs to the WELD METAL, set by the electrode class and your code, and it already carries that code's safety factor. Look it up; do not remember it, and do not apply a second factor on top.