The author recommends that users first measure their own one-week usage, because their own workload determines which lever has the greater impact and they should not simply adopt others' numbers.