Understanding forecasting consumption


In this article we’ll quickly introduce you to the concept of “forecast consumption” (or “forecast netting“). In a few minutes you’ll understand its role in your planning processes.

Customer demand in your supply chain comes in two forms:

  • Sales orders you already have received from your customers.
  • Sales forecasts that are a prediction of the customer orders you expect to sell in a certain period

The importance of each form varies by business. For instance, a grocery shop will plan its replenishments only based on a forecast. An engineer-to-order company will plan only the customer project orders it received.

Many companies operate with a mix of sales orders and forecasts. The short term demand consists mainly or enterily of sales orders. Long term demand consists mainly of forecast. For some period both forecast and sales orders will coexist and this is where forecast consumption comes into the picture.

The need for forecast consumption is obvious when you look at the short term period in the above picture.

Planning the “80 sales orders + 100 forecast” would double-count some demand (since it ignores the fact that sales orders are already anticipated in the forecast).
Instead, we want to plan the “80 sales orders + 20 the remaining part of the forecast not realized yet as sales orders”. The calculation of this “remaining forecast” is called forecast netting or forecasting consumption.
The end result after the forecast netting is a consistent demand signal that we can use to plan across the complete planning horizon.  
All planning decisions can be made using this combined, consistent demand signal. 
We don’t need to separate the short-term plan from the long-term plan, but can combine both in a single plan.
Let’s see how this drives our planning decisions across time:
  • Let’s follow a time bucket 6 months in the future.
  • That far in the future the demand  consists of only forecast of 100.
    This demand will already trigger the procurement of raw materials and components that have a long lead time of 5 or 6 months.
  • Three months go by, and our time bucket is now only 3 months in the future.
    In the mean time we received sales orders for 20 pieces already.
    The plan we generate now is based on 20 sales orders plus 80 net forecast.
    This demand will trigger the procurement of raw materials and components that have a lead time of 3 or 4 months.
  • Another three months pass, and our time bucket is now in the current month.
    We have received sales orders for 90 pieces.
    The plan we generate now is based on 90 sales orders plus 10 net forecast.
    This demand will now trigger the procurement of raw materials and components that have a lead time less than a month.
From the example you can see that long lead time materials are purchased purely based on a forecast. And purchasing of short lead time materials is driven almost entirely by the sales orders. 
Sounds easy and logical, but it’s the concept of forecast consumption that allows us to plan consistently and smoothly across this whole horizon, across a mix of different demand types and different lead times. 
You can read more and see this feature in action in the frepple software on this forecast netting example page.  It covers some more complex cases than we adress in this short post: e.g. what happens if orders exceed forecast in a certain period?   E.g. what happens if the forecast is for week 11, but the customer order is placed for week 10?
Curious about testing forecast netting on your data ? The forecast module of frePPLe is 100% free and open source. You can get frePPLe from the download section and give it a try.

Related articles

How mature is your supply chain planning process?

Planning vs scheduling processes

Backward or forward scheduling?

Demand classification: why forecastibility matters

Why not using an ERP as planning system?

Decoupling points in supply chain

Comparing MRP, Excel, spreadsheets & APS

Timeline of a project