In the previous phase we imported the scenario and examined the statistics with the data gathered by the Simulation experiment. In this phase we will set up and run the variation experiment to define the optimal parameters of the Inventory policy.

Create and configure Variation experiment

1. Navigate to the Experiments section and select Variation experiment. You will be taken to the Variation experiment's view with its default settings.

   

2. Leave the experiment's default start and end dates.
3. Click the Use replications toggle button and specify 3 in the Replications per iteration field.

   We need replications since demand of the Boston customer is set to a probability distribution. The model will have stochastic behavior resulting in a less representative experiment run. To improve representability we will enable replications, which will result in conducting several runs per each iteration. Keep in mind that replications require more time to complete the experiment.

   

4. Click the Product statistics unit drop-down list and select pcs, since it is the product measurement unit defined in the Products table.

   

5. Enable the Lost orders statistics:
   • Click Configure statistics.
   • Type Dropped Orders in the Name column filter field. Press Enter. The result of the search will return the required statistics.

     

   • Click the toggle button to the left of the Demand Placed (Dropped Orders) by Customer to activate data collection for these statistics.
6. In the same way check the Total costs statistics. It should be enabled by default.
7. When done, remove the filter by clicking the icon to the right of the statistics name and type true in the Enabled filter field to see all the enabled statistics.

   

8. Click OK to close the Configure statistics dialog box.

Now we will define the variable parameters, i.e., the data the experiment will work with. In our case we will vary the Min and Max parameters of the inventory policy.

A variable parameter that you define is set for a single parameter only. If you need to vary several parameters during the experiment, you need to create a separate variation for each parameter.

1. Click Add beneath the Variable parameters to open a dialog box with the scenario's parameters that you can vary.

   

2. Define variation of the min parameter as shown on the screenshot below. Click OK to close the dialog box.

   

3. In the same way create variation of the Max parameter and set its Min and Max values to 2000 and 5000 respectively.

   There, we have defined two variable parameters of the inventory policy.

   

We have completed setting up the Variation experiment and may run it now.

1. Click Run. A table will appear updating its records with results received in real-time. Experiment execution will take some time.
2. Perform sequential sorting of the data based on the lost orders column to instantly see the best currently available result:
   • Click the mean name below the Demand Placed (Dropped Orders) by Customer column title to sort the records in the ascending order.
   • Press and hold the Alt key, then click the mean name below the Total cost column title.

   

3. Observe the final results. When the experiment is completed, you will be taken to the Variation results tab of the new Result item of the Variation experiment branch, which contains the results. Sort the data in the same way as you did after starting the experiment.
   As you can see there are four successful results with no dropped orders. the best one is the top one. According to it, if the Min and Max values of the Inventory policy parameters are set to 2000 and 2,600 respectively, we will not lose a single order and we will incur the total expenses of $169,878,910.793, which is more than the initial amount, but it is the price we have to pay to maintain high quality of delivery service.

   

You can navigate to the Inventory table and update the parameters of the defined policy with the acquired data.

The received results can be improved by adjusting the variable parameters for the experiment.

Based on the range of values that we specified for the two variables, the experiment suggests the Min and Max values to be set to 2,000 and 2,600 respectively. Now that we know these Min and Max values, we may adjust the variable parameters by altering the range of the variables and the step of variation to receive a more precise result.

Adjust the variable parameters

1. Select the first variable and click Edit below the Variable parameters box to open.
2. Change the Min and Max values to 1,700 and 2,300 respectively.
3. Set Step to 50.
4. Click OK to close the dialog box.

   

5. In the same way adjust the second variable by changing its Min and Max values to 2,300 and 2,900 respectively, and setting Step to 50.
6. Run the Variation experiment and examine the final results, which will be located in Result 2 item of the Variation experiment branch.

   

7. Compare the received results with the results from the Result item of the Variation experiment branch. You will see that the incurred expenses are less this time and we have still managed to satisfy our customer's demand.

   

The received result can be adjusted again to provide even more precise inventory policy parameters. You can:

1. Decrease the specified range of each variable parameter (e.g. Min — 1800-2200, Max — 2400-2800).
2. Decrease the step size from 50 to 25, 10 or 5 (the smaller the step is, the more time it requires to complete the experiment).
3. Increase the quantity of replications to 5-10 (the more replications per iteration, the more time it requires to complete the experiment).

You can navigate to the Inventory table and update the parameters of the defined policy type with the acquired data.

We have successfully completed the Variation experiment tutorial.

How can we improve this article?

Send