Previous empirical evidence which evaluated the accuracy of management earnings or sales forecasts consistently revealed these forecasts to be on average signi cantly overoptimistic. However, all studies analyzed forecasts from public disclosures, which are an important signal to investors and analysts and thus possibly biased by strategic considerations. To disentagle whether and to which extent strategic deception or cognitive biases are resposible for this overoptimism, the present study analyzes the accuracy of 6,234 undisclosed, company-internal sales forecasts, which German firms provided anonymously to the IAB Establishment Panel. Quite surprisingly, the study reveals the average forecast to be signi cantly overpessimistic. I propose that the non-existence of a general bias towards overoptimism is due to the lack of incentives to consciously overgloss future prospects in undisclosed forecasts and that overpessimism may be a consequence of loss aversion.