I found this article interesting because it shows empirical analysis against the commonly held belief that "errors are more costly the later it's discovered in the software development process" including myself.