If you hold stock at more than one location — a central warehouse feeding regional depots, or raw material at a plant feeding finished goods at distribution points — you have a multi-echelon network. Most manufacturers in this position manage each location as though it stood alone, which is intuitive, straightforward, and systematically wasteful.
Why Independent Management Overstocks
Set safety stock separately at each of five depots and each one buffers against its own demand variability in isolation. But demand spikes at those depots rarely coincide. When one runs high, others typically run normal or low. Sized independently, you’re holding five buffers against five worst cases that will never occur simultaneously.
The statistical term is risk pooling, and the practical effect is substantial. The total buffer a network genuinely needs is materially less than the sum of independently calculated buffers — the more locations you have, the wider that gap becomes.
Where to Hold The Stock
The second question multi-echelon thinking answers is positioning. Stock held centrally serves any location but takes longer to reach the customer. Stock held forward is immediately available but committed to one location, and if demand appears elsewhere it’s stranded.
The right split depends on how predictable local demand is and how quickly you can move stock between tiers. Where local demand is stable and replenishment is slow, push stock forward. Where demand is volatile and you can transfer quickly, hold it central and keep your options open. Most networks are set up on neither principle — stock ends up wherever it landed historically.
The Postponement Idea
A related move is to hold inventory in a less finished state for longer. If four product variants share a common base and differ only in final packaging or configuration, holding the base and finishing to order pools your variability across all four. You need far less total inventory to hit the same service level, because you’re no longer betting on the variant split.
This isn’t always feasible — it depends on whether the differentiating step can be done quickly and close to the customer. Where it is feasible, it’s usually the largest single inventory reduction available, and it’s a supply chain design decision rather than a planning one.
What it Takes to Run
Multi-echelon optimisation needs demand and lead time data at every node, plus transfer times and costs between them. The calculation itself is beyond spreadsheet work, which is the honest reason most manufacturers don’t do it rather than any disagreement with the logic.
It also requires that the network be managed centrally, at least for stocking policy. If each depot manager sets their own levels to protect their own service metrics, you’ll get independent buffers regardless of what the model recommends. The organisational change is usually harder than the analytical one.
A reasonable first step is to quantify the prize. Sum your current safety stock across locations for one product family, then calculate what a pooled network would need at the same service level. The gap is your opportunity, and it’s usually large enough to justify the organisational conversation that follows.
ticktick.ai calculates stock levels across the full network rather than location by location, accounting for pooling and transfer options.
