Why does the way we naturally measure complexity stop working when we apply it to programming? Lots of projects that have nothing to do with programming end up

Why does the way we naturally measure complexity stop working when we apply it to programming? Lots of projects that have nothing to do with programming end up taking longer than expected or are over budget.