Learning from a mistake in a shear test