I know… the title sounds like a contradiction, but in this case there are two forecasts involved: one in the DC and one in front of the bottleneck.
We are currently redesigning a supply chain for a process manufacturer and one of the problems is, that we have to be careful so that the bottleneck work center, with its limited capacity, always is being fed the right job. Therefore we avoid making too much of that stuff that doesn’t sell and avoid ending up having no capacity to make that stuff that does sell.
Imagine a line that produces bottled fruit juices. In the raw materials warehouse we have apples, bananas, oranges and mango (there is sugar and preservatives too and empty bottles, caps and labels). The first work center is a press which takes the fruit and out comes nectar being put in inventory. After that, the next station is a blending tank, taking one or more nectars, sugar, water and some other stuff to produce the finished juice (banana, orange, apple or tropical punch). That blending tank is the bottleneck! The juice is stored in barrels and then gets bottled and shipped to the distribution center in cases on pallets.
My suggestion was: Put a forecast for each individual juice into the DC (strategy 40) and have the planning run generate stock transport orders to create demand for the line. The finished products in the bottling plant are set up for repetitive manufacturing with production versions, so that the bottling schedule is automated and takt or rate based sequencing.
In front of bottling, there are Kanban containers with finished juice – in a simple scenario there are two barrels of each juice. If you take a barrel and make 10 pallets of bottled juice, the Kanban is set empty and a process order is generated to make another barrel of that juice. Therefore that process order is pulling a specific nectar (or more) with sugar through the bottleneck and makes sure that the bottleneck is loaded ONLY with something that goes out at the end.
Now there must be enough nectar of the specific fruit available and for that I suggested a (seasonal) forecast on component level (strategy 70) which also gave a secondary requirement onto the long lead time raw materials
My client does not make fruit juice but they are a process manufacturer and Kanban and takt based scheduling works very well for them. There are many, many different specs and the guesswork on what spec to push through the oven has ended. The operators are now loading according to a signal from a kanban container… and they know exactly what to load onto the capacity-limited line.