Latest Blog Entries
Archives
 

21 Jul Forecast Consumption Case Study, Part II: A Comedy of Errors

In my last blog, (click to read part 1 of this 2 part series), I outlined a few real life errors of a previous client to demonstrate how multiple, seemingly logical singular decisions in the forecast consumption arena could compound on one another to achieve some really ugly results. In this blog, I am going to use an example to show how these decisions, given a 100% accurate forecast (from a sales/demand planning perspective), can cause serious problems for your supply team, and make your demand planners look responsible.

Example

Monthly forecast

The sales person relays to the demand planner, during the monthly sales meeting, that the customer intends to buy 10,000 cases of product on July 27th. The demand planner enters a forecast of 10,000 CS for the month of July in demand planning (remember that, in this case, the demand plan is entered in monthly buckets only).

Forecast Release and Week 1

The forecast is released to supply planning and split into 4 weekly buckets (assuming even buckets for easy math), resulting in forecast requirement orders of 2,500 CS for each week in July. In response to the demand in week 1, a production order is created and executed, resulting in 2,500 CS of inventory in week 1.

W1

Week 2

The unconsumed forecast from week 1 was deleted, which results in the planning system no longer needing the production order in week 2, as the inventory in week 1 will cover week 2’s requirement:

W2

Week 3

The unconsumed forecast from week 2 was deleted, which results in the planning system no longer needing the production order in week 3 as the inventory in week 1 will cover week 3’s requirement:

W3

Week 4

And now the fun begins… In the middle of week 4 (after week 3’s forecast requirement had been deleted) the sales that had been predicted is entered. Since there is no forward consumption (remember that we wiped it from the collective corporate memory) and the forecast requirements in the past have been deleted, the sales order can only consume the requirement in week 4. This result (from the view of a supply planner) can misleadingly look as though the item was grossly under forecast and in this case, caused the need for 7,500 cases to be produced in a period where only 2,500 CS were expected and the supply planners had to scramble to be able to fill the order (assuming that the raw materials purchases weren’t delayed due to the perceived lack in demand).

 W4

5 Lessons Learned from this Case Study

Your forecast strategy (mode and days +/-) should be analyzed and set based on the demand volatility.

Your forecast strategy should be re-analyzed on a regular basis, usually on about the same timeframe as statistical forecast models are re-evaluated.

Splitting a monthly forecast into weekly buckets is fine, but make sure that your forecast strategy can keep the integrity of the intended initial forecast. If possible, align your demand planning organization to use the same time buckets as the supply organization.

Make sure that your forecast requirement deletion strategy is in line with the forecast consumption settings. If you set a backward consumption period of 30 days and then delete the forecast requirement every week, you have effectively reduced the backward consumption period to 7 days

Lastly, DO NOT SIMPLY SET AND FORGET! Monitor how well the forecast is being consumed and adjust those settings accordingly. Consider creating an alert that will notify demand/supply planners when the forecast is being grossly over/under consumed.

You can also read part 1 of this 2 part series here.

3 Reasons Why APO is Outdated and 1 Why it is Not
Forecast Consumption Case Study, Part I: A Comedy of Errors