Multi-echelon inventory: setting stock across levels together

Independent settings overstate what the network needs

Each node calculating its own buffer from its own demand history assumes it must survive unsupported. In practice an upstream location can resupply a node that runs short, so part of the protection each site holds is duplicated across the network. Setting the levels jointly recognises that shared cover and allows the same availability to be delivered with less capital, provided the internal replenishment between levels is fast and reliable enough to be relied upon.

Pooling favours holding uncertainty upstream

Demand aggregated across several downstream points varies proportionally less than each point does individually, so a given quantity of buffer protects more effectively when held centrally. This argues for thin local cover sized to the internal replenishment interval, backed by a pooled upstream buffer sized to supplier variability. The counterweight is delivery speed, since stock held centrally is further from the customer, which is why the split depends on the service promise for each segment.

The inputs that determine the split

The decision turns on the variability of demand at each node, the lead time and reliability of the internal link between levels, the lead time and reliability of external supply into the top of the network, the service target at the customer-facing level, and the relative cost of holding stock in each form and location. Where the internal link is slow or unreliable, the pooling benefit shrinks, and the network is pushed back towards local protection.

Implementation ahead of optimisation

The mathematics of joint optimisation is well established and is not usually the obstacle. What blocks results is unreliable stock records, internal replenishment that misses its own lead times, local managers who quietly hold extra cover because they do not trust the upstream position, and measurement that penalises a site for a stockout while ignoring network stock. Fixing the reliability of the internal link, and measuring availability at the network level, delivers much of the benefit before any advanced tooling is involved.

Frequently asked questions

Does this approach always reduce total stock?

It reduces it where downstream nodes can genuinely rely on upstream cover. Where the internal link between levels is slow or unreliable, the calculated saving evaporates in practice and local sites rebuild their own protection informally, which is worse than planning it openly.

Why do local sites keep hidden buffers?

Because they are measured on their own availability and have been let down before. The remedy is to make the internal replenishment dependable and to measure service at the level the customer experiences, so protecting the network no longer requires each site to protect itself.

How does this differ from ordinary safety stock calculation?

Ordinary calculation sizes a buffer for one location against its own uncertainty. The multi-echelon view decides how uncertainty should be distributed between locations first, and only then sizes each buffer to the role its level has been given.