Sales forecasting updates: how good are they in practice?

Both theory and the opinions of practising forecasters expect that, as the lead time before an event reduces, updates to the forecast will efficiently incorporate more recent information and so be more accurate and less biased. However, in practice, it may be anticipated that there will be some inefficiency due to excessive anchoring on the last forecast, leading to positive correlations in forecast revisions. This study tests these expectations using a large sample drawn from judgementally estimated sales forecasts from 10 manufacturing organisations. The results suggest that forecast accuracy does not improve as much as anticipated as the lead time reduces, and that the forecast revisions display negative not positive first-order autocorrelations. The inefficiency of the fixed-event forecast revisions does not appear to be related to the rolling-event forecast accuracy. This is in distinction to the one period ahead forecast errors where efficiency was strongly related to forecast accuracy performance. Some reasons for the findings are discussed.