Forecast Netting - The neglected stepchild of planning
The key link between demand and supply planning is forecast netting. The demand plan passes the gross forecast to the supply plan. Before it can be consumed to generate the supply, it must go through a netting process. The sales orders must be deducted from the forecast to generate the netted forecast. This may sound simple at first but it can get quite tricky. A bad forecast netting process can quickly result in inventory overages and shortages down the line.
The challenge with forecast netting is due to the misalignment between the forecast and sales order granularity. To explain this problem, let's first look at the dimensions of the demand - namely product, customer, and time. Each of these dimensions has a hierarchy. The product hierarchy starts with all products and ends with the Stock Keeping Unit (SKU). You can have multiple layers in between representing brands, product categories, styles, etc. Customer hierarchy starts with all customers and ends with a customer ship-to address. Similar to a product, you can have several layers in between such as channels, customer groups, etc. Finally, the time hierarchy starts with all time and ends with a date. You can roll it up through weeks, months, quarters, years, etc.
In most cases, when a demand plan generates a forecast, it is not at the granular level. Forecast at the SKU, ship-to and date level will be terribly wrong. On the other hand, if the forecast is at the highest level (all customers, all products, all-time), we cannot do any supply plan against it either. Thus, companies pick middle layers that reduce their forecast error while providing sufficient information to plan the supply. For instance, the demand plan may be at style, channel, month level. Let's compare this to the sales order level which we need to net against. Sales order is at the lowest SKU, ship-to, and date level. This is the misalignment I highlighted before.
At this point, we have several choices. We can either dis-aggregate the forecast from a higher level to a lower level or aggregate the sales order from a lower level to a higher level. We can also pick different methods for different dimensions. Our choices will have an impact on the supply plan resulting in inventory shortages and overages. For instance, we can net the forecast at style color, customer group, and week level. To do this, we need to aggregate the sales orders from SKU to style-color, ship-to to a customer group, and date to week. On the other hand, we need to dis-aggregate the forecast from style to style-color, channel to a customer group, and month to week. Aggregation can be done easily by a summation of the sales orders. Disaggregation must use some ratios to bring the forecast from higher levels to lower levels. These ratios are generally calculated using the sales order history.
Things can get even trickier when you have more sales orders than your gross forecast in a time bucket. The key decision is what to do with the excess orders. Should you jump into another time bucket and consume its gross forecast? Should that bucket be prior to or after your current bucket? or should you just ignore the excess?
If you are interested in learning more, please connect with me on LinkedIn, follow me on Twitter, or watch me on YouTube.
My name is Cem and this has been another gem.
