why some software testing automation projects failed while others succeeded?