So, that's it for session 3 of week 4.
How did Idea's problem change from last session to this one?
As before, we simulated the outcomes for the weak and strong markets.
The demand model had a 50/50 chance that the market would be weak or strong.
For each case, we then simulated uniformly distributed demand.
This time, however, the structure of the decision problem became more complex.
First, IDEA needed to decide on a supplier, either S or P, or no one.
For supplier P, I could then decide on an order quantity.
Rather than running a separate simulation for each possible Q,
and we could have run a separate simulation for Q equals 4000,
Q equals 4001, Q equals 4002,
we used a common set of simulated demands for all the possible Q's.
And we then optimized to find an "approximately optimal" Q.
In fact, we essentially solved Senthil's newsvendor problem from Week 1.
And in next session, we'll go back to see how we do it.
In this session,
we extended our model of decision making to include more complex decisions.
We simulated a single set of demands and
calculated average profit as a function of them, and of IDEA's order quantity Q.
By optimizing the spreadsheet,
we estimated the Q that would maximize IDEA's expected profits.
So now we've seen how we can use three analytics tools together to evaluate
potentially complex decisions to be made under uncertainty, decision trees,
simulation and optimization.
Not only that, we also know how to crack the news vendor problem that Senhtil
introduced in week one of the course.
In session 4, we'll go back to the news vendor to see how it's done.