I have found a solved example of A Stochastic Two-Period Model with No Setup Cost in the book Operational Research by Hillier, 7th edition, that has a lot of complicated calculations to arrive to the solution.

In the following example, $c=$cost of ordering, $h=$cost of holding, $p=$shortage cost and $𝑦^0_𝑖$ is the optimal order-up-to level in period 𝑖.

enter image description here

The part where things start to complicate, is when making the substitution of $y_1^0$ into $C_1(x_1)$.

Clearly will not be an easy calculation by hand since $C_1$ has this function $L$, that it's evaluated with diferent arguments, which are two integrals.

I am not sure what form the argument of $\displaystyle \min_{y_1\ge x_1}\{\dots\}$ will have, nor what would be best to do next. My book mentions this OR Courseware but since I have the pdf file book I don't have the disk? is refering to.

And my question is what can I do here in order to find the optimal $y_1^0$ without having to deal with lots and lots of calculations by hand?

Any suggestions are very appreciated.

enter image description here

The $C_1(x_1):$ enter image description here

The $L(z):$ enter image description here


Following the comment of Larry, I substitute $y_1^0$ in the equation equal to zero (below the $C_1(x_1)$ definition in the image).

I assumed $y_1^0=5$ to be the optimal and this is what I got

$$-15+(15+10)\frac{5}{10}+(10-15)\Phi(5-2)+(15+10)\int_0^{3}\Phi(5-\xi)\phi_D(\xi)d\xi $$ $$=more\ calculations$$ $$=-\frac {11}{8}$$ This is different than $0$, which I don't understand why, the equation should have had the value $0$ because the book mentioned the optimal was $y_1^0=5$ and not $y_1^0=6$. Also the optimal should satisfy the equation equal to zero, but does not.

I really don't see what am I doing wrong? I did check twice the calculations of the expression and found no errors.

$\star$ Note: I asked this question on MSE https://math.stackexchange.com/questions/3246019/why-the-optimal-value-that-minimizes-a-function-does-not-satisfy-condition and currently has a bounty (grape period bounty to be exact) and I asked to migrate but it was not possible :(

  • 1
    $\begingroup$ This question is very hard to follow. You haven't defined the notation (e.g., $y_1^0$) or provided sufficient context. The actual questions you are trying to ask are sprinkled throughout the post and are not that clear. I would encourage you to edit the question to make the question more concise and easy to follow. Nevertheless, I will try to answer and see whether that helps. $\endgroup$ Jun 11, 2019 at 2:47
  • $\begingroup$ @LarrySnyder610 Hi Larry, the context (the $y^0_1,y^0_2$) is the same as the other question of mine that you answered. I'll edit anyway. $\endgroup$
    – user441848
    Jun 11, 2019 at 3:02
  • 1
    $\begingroup$ I know that. But other readers of this question will not know that. You didn't even link to the original question. Remember that SE sites are supposed to be building a body of knowledge that will potentially be of use to many people, not just the individual person asking the question. $\endgroup$ Jun 11, 2019 at 3:12

1 Answer 1


The text from the book tells you that the optimal value of $y_1$ is $y_1^0 = 5.42$. This comes from solving the optimality condition (the equation after "satisfies the equation") for $y_1^0$.

Presumably, somewhere in the example it says that the base-stock levels must be integers. Therefore, they plug the neighboring values, $y_1^0 = 5$ and $y_1^0 = 6$, into the objective function to see which one gives the better value. (This relies on the objective function being convex.) They found that $y_1^0=5$ is better. But it is not surprising that when you plug $y_1^0=5$ into the optimality condition, you don't get $0$; $y_1^0 = 5$ is only optimal because it has to be an integer.

(Similarly, if you want to minimize $(x-\frac14)^2$, the optimality condition is $2(x-\frac14)=0$; if $x$ must be an integer, you'd plug in $x=0$ and $x=1$ and pick the better value; but if you plug that value into the optimality condition, the LHS won't equal $0$.)

Now, to calculate the objective function for $y_1^0=5$ or $6$, I think they (or you) are skipping some logic here. The optimal policy is a base-stock policy, which means that you are looking for the optimal base-stock level, $y_1^0$. The cost of a given base-stock level $y_1$ is not given by calculating $C_1(y_1)$; it is given by setting $x_1 = $ something small (e.g., $-10$) and then calculating what's inside the $\min_{y_1\ge x_1}\{\cdots\}$ for that value of $y_1$. The idea is, assuming the inventory level $x_1$ is small enough that we will place an order in this period, what will be the cost if we place an order up to $y_1$?

The line in the book "substituting $y_1^0=5$ and $y_1^0=6$ into $C_1(x_1)$ seems misleading to me because it sounds like you're supposed to set $x_1=5$ and $6$, but that is wrong.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.