I totally agree that this dramatically underestimates the cost of a bad hire. It's more accurate for a hire not hitting expected performance, but in my experie

I totally agree that this dramatically underestimates the cost of a bad hire. It's more accurate for a hire not hitting expected performance, but in my experience a bad hire has a much more extreme worst case. This is especially true in programming because the effect of an individual's decisions gets amplified by large numbers in many cases (the number of internal and external customers downstream of these decisions).