This month there was an article in the Harvard Business Review by Bart de Langhe, Stefano Puntoni (of RSM) and Richard Larrick about estimates of errors. Intuitively, we usually assume that all relationships are linear, while most are nonlinear. Here, the discussion is continued.