Dr. Watson Says: |
 |
| ...Create a robust design that holds up well under a variety of changes to the data. |
|
What Do You Say?
|
|
|
|
At the start of 2013, I wrote an article on the five models you should build this year. It is not too late, the models will still help you.
I just now saw a question in the comment section from Richard Schmidt of the Keystone Business Group:
“Upon completion of the modeling, what are the ways to compensate for the lag time in information that may result in increasing the level of risk into the supply chain?”
This is a good question. If you are doing a serious modeling exercise, say to change the number and locations of your warehouse or plants, the project could take three to four months to complete. It may take two more months to get all the approvals that you need to implement your solution. And, maybe you don’t start breaking ground on the new facilities for another six months.
If you started the project with data from the previous twelve months, by the time you break ground, some of the data you used is now two years old. Is there some risk that you are making the wrong decision because the data is no longer fresh enough? Sure. A lot can change in two years.
There are several strategies for mitigating this risk.
But, first let’s discuss what you shouldn’t do: You shouldn’t continually update your data through the analysis process. At some point, you lock in the data with the best information you have at that time. You then do your analysis on this set of data. If you try to change your data, you will burn all your project time just changing the data and not actually analyzing the supply chain. You won’t be able to come to a decision.
Here are the three things you should do:
1. Create a robust design that holds up well under a variety of changes to the data. Rather than creating one perfect solution for the single version of data you start with, you would to search for a good solution that holds up well to changes in your data. For example, test what happens if demand goes up or down or transportation costs go up or down.
2. Create multiple acceptable solutions in the design phase. Then, based on updated data or new insights, test each of your designs again right before you have to make a decision. This way, the final test is relatively quick (because you’ve narrowed your choices) and your data is as fresh as it can be.
3. Create a time-phased implementation plan. In complex models, your final solution may show several different major changes. A time phased plan shows which changes you will make this quarter, which ones next quarter, and so on. You want to create time phase the changes so that you do the ones with the most savings first or those that are easiest to implement. Then, the key idea is that re-run the model before implementing the next phase to see if you need to make adjustments to the plan.
Final Thoughts
It should be noted that these steps are part of a larger theme that we frequently discussion this column and on this website: you should have modeling capability in-house to continually model your supply chain. Supply chain modeling is not a one-time activity.
|
