The newsvendor problem is one of the classical operations research
models. It can be found in almost any operations management and inventory
textbook. It is a single-period, single-product inventory problem, with a
single procurement before a stochastic demand realization. Several empirical
studies have shown that students and professional buyers do not order a
quantity similar to that proposed by the classical solution of the newsvendor
problem, which has the objective function of maximizing the expected profit.
There are many possible explanations. This research was motivated by a desire
to explain the deviation from the expected profit maximizing order quantity
with one simple decision making criterion. Furthermore, we explored a new
decision paradigm. It is to this end that we studied the newsvendor problem
with the info-gap decision methodology. The info-gap theory has been shown, in
some cases, to explain the behavior of animals better than other decision
making methodologies. Our study compares, both quantitatively and
qualitatively, the solutions obtained by the classical, maximum probability of
achieving a critical profit level and Info-Gap methodologies. In our analysis
of the newsvendor problem with the basic Info-Gap formulation we do not assume
any probabilistic demand distribution. We only use an estimate of the nominal
demand and the error of this estimate. We do not assume knowledge of a worst
case and do not perform a max-min analysis. Info-Gap decision theory allows us
to determine the robustness of the decision to the uncertainty, namely, how
much the demand can deviate from the nominal quantity and we can still assure
achieving the critical profit. A critical profit is an additional input
required for the analysis with info-gap decision theory. We successfully
explain the empirical results in some cases, whereas in other cases there is
more work to be done. Two additional analyses include a hybrid model that
incorporates uncertain probabilistic information in an info-gap analysis and a multi-item
model.
