Rapid advances in information technology have brought decision makers
the mixed blessing of an increasingly vast amount of easily available
data. Designers of decision support systems (DSS) have focused on inco
rporating the latest technology with little attention to whether these
new systems are compatible with the psychology of decision makers. Ou
r premise is that DSS should be designed to take advantage of the dist
inctive competencies of decision makers while using technology to comp
ensate for their inherent weaknesses. In this study we apply this appr
oach to a forecasting task. We find that to arrive at a forecast decis
ion makers often search their experience for a situation similar to th
e one at hand and then make small adjustments to this previous situati
on. Our theoretical model of the performance of this intuitively appea
ling strategy shows that it performs reasonably well in highly predict
able environments, but performs quite poorly in less predictable envir
onments. Results from an experiment confirm these predictions and show
that providing decision makers with a simple linear model in combinat
ion with a computerized database of historical cases improves performa
nce significantly. We conclude by discussing how these results can be
used to help improve forecasting in applied contexts, such as promotio
n forecasting in the retail grocery industry.