Forecast Reconciliation

This is why it is important to understand what the right level is to develop the forecast models. Partially, this question depends on what the customer wants.

Customers may want a forecast at the detail level - especially the supply chain planning folks. But sometimes may be not. So it is important to understand how the supply chain is structured in order to decide on the forecast delivery level, a level at which the demand planning group has to deliver the forecast to supply planning group.

So it is the responsibility of the demand planning group to deliver forecasts at the agreed upon level to the supply chain. At most companies, for most supply chains, this happens to be the SKU level. And in some companies the SKU may coincide with a pack level and at some companies may be not. It depends on how your supply chain is structured.

If the lead time to produce the finished product is much longer say 30 days, than the time it takes to pack out and ship say half a day, then perhaps the supply chain may decide to carry inventory at the finished product and NOT at the pack level. It’s probably a good idea to forecast and deliver on SKU level, but the pack forecast will be determined on the base of allocation, as more of an execution strategy or this could be even on demand. So the forecast level is the level of your supply chain bottleneck or the level at which you carry the most inventory for on-demand customer fill.

This is where the question of what is on demand and what is in inventory plays a big role. If the product can be packed on demand, then inventories could be kept at SKU level, and forecast could be delivered at SKU level, which also means you could get more accurate forecasts. Remember, the higher the level in the product hierarchy, the more accurate the forecast models are.

Getting back to forecast reconciliation, it is important to distinguish between forecast modeling and forecast delivery. Within demand planning itself, it’s quite possible to model at a particular level but then disaggregate and come up with a forecast at a more detailed level and deliver a forecast at that level. The middle-out forecasting method is particularly useful since you develop the forecasts at a middle out level in the hierarchy but deliver a detailed forecast for the supply chain using disaggregation methods and aggregate the forecasts for financial planning at the higher level.