This paper discusses the model for a perishable product 

with two types of states, a good state and a bad one. 

The commodities sell well when the good state occurs 

at the beginning of a period, but do not when the bad 

state occurs. 

An indicator random variable is defined to express 

the states, and it is a Bernoulli random variable 

with unknown parameter which has a conjugate beta prior. 

We express the maximum expected profit for remaining periods 

by a dynamic program and obtain the optimal ordering policy 

in the first period. 

As a main result, we get a relationship between the numbers 

of times at which the good and the bad state occurred and 

the expected profit for remaining periods. 



Moreover, numerical examples are given to illustrate an 

optimal ordering policy for remaining $n>1$ periods.