INVInventoryStop 8 of 11
Multi-echelon inventory (MEIO)
Placing buffers across plant, warehouse and store as one system, instead of letting every site protect itself.
Read atChange any time
Most products pass through several stages, or echelons, before a customer buys them: plant, central warehouse, regional warehouse, store. If each stage sizes its safety stock on its own, every stage protects itself against the same uncertainty. The result is buffers stacked on buffers.
Multi-echelon inventory optimization (MEIO) treats the whole chain as one system and asks: where should the buffers sit so the end customer gets the target service at the lowest total inventory?
Think of a shampoo network: one plant, a central warehouse and three regional warehouses. Two very different set-ups can give customers the same service:
- Push stock forward, holding most of it in the regional warehouses close to customers. Delivery is fast, but each region carries its own buffer.
- Hold it centrally, keeping a larger pooled buffer at the central warehouse and leaner stock in the regions, which are refilled quickly from the centre.
MEIO weighs these options using each stage’s lead time, demand uncertainty, holding cost and the service promise to the next stage. A common outcome is less stock at the expensive, fragmented end of the network and more at a pooled point upstream, though fast-delivery promises can push stock forward again.
That’s why MEIO appears in planning software as a separate capability from single-site safety stock. It needs good data across the network and a clear service target per customer segment.
Words to use
- Echelon
- One stage or tier of the network: plant, central DC, regional DC, store.
- Service time
- The delivery time one stage promises the next; longer promised times let downstream stages hold less.
- Guaranteed-service model
- A classic MEIO approach that sets each stage’s promised service time and sizes buffers to meet it.
Each stage quotes a service time
The approach most used in practice, the guaranteed-service model, treats every stage as a promise. Each stage quotes the next stage a service time: how long an order will take to ship. A stage then needs safety stock only for its net replenishment time:
If a stage promises its customer the full time it takes to be resupplied, its net time is zero and it holds no buffer; it simply passes orders through. If it promises immediate shipment, it must cover its whole replenishment time. The optimizer chooses these internal promises to minimize total holding cost while keeping the external promise to customers.
A central warehouse (CDC) is resupplied by the plant in 9 days. Regional warehouses (RDCs) are 2 days from the CDC and must ship to customers immediately. Each region’s daily demand has = 100 units, = 1.65, and holding a unit costs 1.0 at the CDC and 1.2 in a region.
One region:
- A, the CDC promises immediate shipment: CDC buffer = 495, plus an RDC buffer for 2 days of 233. Total 728 units, cost 775.
- B, the CDC passes orders through (it promises 9 days): the RDC covers 9 + 2 = 11 days, 547 units, cost 657. B wins: one buffer over 11 days beats two separate buffers, because .
Three independent regions:
- A: the CDC pools all three () with a buffer of 857, plus 3 × 233 in the regions. Total 1,557 units, cost 1,697.
- B: each region covers 11 days on its own, 3 × 547 = 1,642 units, cost 1,970. A wins: pooling three regions at the cheaper central site now outweighs the extra layer.
So “where should stock sit?” has no general answer. It depends on lead times, how many locations each stage pools, and how holding cost rises along the chain, which is why the decision needs a model.
What a MEIO model needs
- The network: stages, links, lead times and their variability.
- Demand and forecast error where customers order.
- Holding cost at each stage, which usually rises as value is added and goods move closer to customers.
- Service promises to external customers, by segment.
Two families of models
- Stochastic-service models (Clark and Scarf, 1960, and the METRIC model for military spare parts) let stockouts upstream delay the stages below and compute the resulting waits. They are exact for some network shapes but hard to scale.
- Guaranteed-service models (Simpson, 1958; Graves and Willems, 2000) assume each stage always meets its quoted service time, covering extreme demand by expediting or overtime. They scale to large networks and handle assembly structures, which is why most planning software uses them or a blend of both.
Internal service can be lower than external service
A counterintuitive result: an upstream warehouse serving the company’s own regional warehouses often should not run at 98% fill rate, even if customers get 98%. Its customers carry their own buffers, so an occasional delay upstream is absorbed downstream. Holding 98% at every stage is exactly the stacked-buffer problem MEIO removes. Internal targets are an output of the optimization, not an input.
Where buffers tend to end up
Typical patterns, which the model confirms or overturns for each network:
- At pooling points, where many downstream locations or products draw on one stock.
- Just before customization, where a generic item becomes many variants.
- At long or unreliable lead times, to protect the rest of the chain from them.
- Close to customers only where the promised delivery time is shorter than the time to resupply.
Making it work in a real company
The mathematics is the easy part. The harder problems are data (lead-time variability is rarely measured well), incentives and execution:
- Local KPIs: if each warehouse manager is judged on their own fill rate, the central site will resist a lower internal target. Measure service where the customer feels it.
- Policy, not just numbers: replenishment rules must actually use the MEIO targets, or planners will override them.
- Re-run regularly: demand, lead times and the network change, and so do the right buffer positions.
Interview kit
Practise it out loud
Answer each question yourself before opening the model answer.
MEIO optimizes inventory across all echelons together rather than site by site. Because stock at an upstream stage protects everything downstream of it, planning each site alone double-counts uncertainty. MEIO decides where buffers should sit, considering lead times, variability, holding costs and the service each stage promises the next, to hit customer service at the lowest total inventory.
What is the difference between single-echelon safety stock and MEIO?
Single-echelon sizes the buffer for one location as if it were alone. MEIO sizes all locations jointly, so an upstream buffer can cover several downstream sites and stock isn’t duplicated along the chain.
Where would MEIO typically move inventory, and why?
Often upstream to a pooled central point, where one buffer protects many downstream locations, while downstream sites hold lean cycle stock. If customers need very fast delivery, MEIO may move stock forward instead. The answer depends on lead times, costs and service targets.
Connected stops
Where this idea feeds in, or gets its inputs from.
Sources and further reading
- Clark, A. J. & Scarf, H. (1960). Optimal policies for a multi-echelon inventory problem. Management Science, 6(4), 475–490.
- Graves, S. C. & Willems, S. P. (2000). Optimizing strategic safety stock placement in supply chains. Manufacturing & Service Operations Management, 2(1), 68–83.
- Silver, E. A., Pyke, D. F. & Thomas, D. J. (2017). Inventory and Production Management in Supply Chains (4th ed.). CRC Press.
- Simpson, K. F. (1958). In-process inventories. Operations Research, 6(6), 863–873.