Buffer stock looks like waste on a balance sheet and like oxygen on a shop floor. The argument is usually won by whoever brings the arithmetic.
Purchasing is measured on inventory turns; production is measured on output. When those two metrics collide over a buffer of subassemblies, the discussion is rarely about the actual numbers, which is a shame, because the actual numbers are usually decisive.
The cost of a buffer is knowable: the units, the capital tied up, the storage, and the risk that a revision change strands them. The cost of a line stop is also knowable — lost output, the labour standing idle, the expedite fees, and the customer-facing delivery slip — but it is spread across departments, so nobody owns it and it rarely appears in the same spreadsheet.
Put both on one page and the answer is usually obvious in one direction or the other. A cheap, stable subassembly with a long lead time is worth buffering. An expensive, revision-sensitive module with a short lead time is not; that one needs a supplier who can deliver reliably instead.
The interesting cases sit between those extremes, and that is where a supplier relationship earns its keep: agreed call-off windows, a small finished-goods buffer held at the supplier, and a rolling forecast that lets the material be bought before it is urgent. The buffer still exists — it just sits where it is cheapest to hold.
— GANI Manufacturing engineering team