Multi-echelon inventory optimization (MEIO) sets stock targets for every level of a supply network at the same time, for example a central DC, regional warehouses and stores, instead of planning each location on its own. It decides how much buffer stock the network needs and where to hold it, so end customers get the target service level for the least total inventory. It matters wherever a DC sits between suppliers and the point of sale, because DC stock changes the risk at every store it serves.
What Is an Echelon?
An echelon is one tier of the supply network. A typical retail network has suppliers, one or more central DCs, sometimes regional DCs, and stores or e-commerce fulfillment centers. A typical CPG network has plants or copackers, a central warehouse, regional DCs and customer delivery points.
Each echelon holds inventory for a different reason. Plants hold cycle stock from production runs. DCs hold stock to cover long supplier lead times and to serve many downstream locations. Stores hold presentation stock and a buffer against daily sales variation. MEIO decides how these buffers work together.
Single-Echelon vs Multi-Echelon Planning
In single-echelon planning, each location sets its own safety stock from its own demand and the lead time it is quoted by its supplier. The store assumes the DC always has stock. The DC treats store orders as its demand and buffers against their variability. Each node may be well tuned for itself, and the network still carries too much inventory, in the wrong places.
Three problems follow:
- Duplicated buffers. The DC and the stores each hold safety stock against the same customer demand uncertainty.
- Hidden lead time risk. When the DC runs short, the store’s real lead time is longer than the planned one, so stores stock out even though their own targets looked right.
- Amplified demand signals. Store orders are lumpier than store sales because of order cycles and pack sizes, and a DC that buffers against order variability holds more than end-customer demand requires.
Multi-echelon planning treats the network as one system. It accounts for the fact that the service the DC gives the stores sets the stores’ effective lead time, and it chooses DC and store targets together. The output is a target for every SKU at every location, with the end-customer service level as the constraint and total network inventory investment as the cost to minimize.
The same issue often has an organizational root. When purchasing owns the DC and store operations owns replenishment, each team optimizes its own stock and its own KPI. A single model gives both teams targets that add up.
Risk Pooling: Why Central Stock Goes Further
Risk pooling is a large part of why MEIO saves inventory. Demand at one store is more variable, relative to its average, than demand across many stores combined, because highs at one location tend to offset lows at another.
A simple example: four stores each sell an item with the same average demand and a standard deviation of 50 units over the lead time, and their demand is independent. Buffering at each store means four safety stocks sized on 50 units each, so buffers sized on a combined 200 units. Buffering the same demand at one DC means sizing on the standard deviation of the combined demand, which is 50 times the square root of 4, or 100 units. At the same service factor, the pooled buffer is half the size.
In practice pooling saves less than this, because store demand is partly correlated and stores still need stock on the shelf. So MEIO rarely puts all the stock at the DC or all of it in stores; it sets the split item by item. Items with high, steady demand tend to keep more stock forward in stores; slow and erratic items tend to keep more of their buffer at the DC, where pooling does most for them.
Multi-Echelon Inventory Optimization in Retail and CPG
Retail DC and stores. A fashion or grocery retailer buys from suppliers with lead times of weeks, or months for imported goods, and replenishes stores from a central DC several times a week. With MEIO, the DC holds the pooled buffer against supplier lead time and forecast error, and stores hold presentation minimums plus a small buffer for the short DC-to-store lead time. The DC’s own service target to stores does not have to be 99%. The model trades a slightly lower DC service level against slightly larger store buffers wherever that lowers total stock for the same shelf availability. Shelf life adds a limit for fresh and chilled items: buffers that would expire before they sell are not an option, so availability is weighed against waste.
CPG plant and DCs. A manufacturer produces in campaigns at a plant, ships to a central warehouse and then to regional DCs that serve retailer DCs. Production cycles create large cycle stocks at the top of the network, and regional DCs need enough stock to meet customer service agreements on short delivery windows. MEIO decides how much of the safety stock sits at the central warehouse, where it covers all regions, and how much sits at each regional DC, where it shortens delivery. Seasonal products add a pre-build question, since capacity limits mean stock for the peak has to be produced and positioned ahead of it.
In both cases the targets feed distribution planning, which turns them into replenishment orders and shipments between echelons. For how inventory targets connect to the service and cost trade-off, see our post on the inventory service level trade-off.
What Data Does MEIO Need?
A multi-echelon model needs the following inputs, at SKU and location level where it applies:
- Network structure: locations, the lanes between them and which location supplies which.
- Lead times by lane, and their variability, measured from receipts rather than taken from master data.
- Demand history at the customer-facing locations, and the forecast error for the planning horizon.
- Service targets by segment, and the measure they use (fill rate or cycle service level).
- Costs: holding cost, order or setup cost and transportation cost between echelons.
- Constraints: order multiples, pack sizes, minimum order quantities, storage and handling capacity, shelf life and presentation minimums.
- Current inventory positions, including stock in transit, and supplier delivery performance.
Data is rarely complete at the start. Missing lead time distributions or service targets can be estimated and refined, as long as the model flags where it made the estimate. Where to put a DC in the first place is a separate decision, made in supply chain network design.
How Solvoyo Runs Multi-Echelon Optimization
Solvoyo’s inventory planning solution uses MEIO to set targets across suppliers, DCs, regional warehouses and stores in one model. Safety stock is sized from actual forecast error, lead time variability and supply uncertainty, and when stock is repositioned at one node, the effects upstream and downstream are recalculated. Distribution then follows from the same targets: demand-driven distribution requirements planning calculates net requirements at every node at once, taking transit times, minimum shipment quantities, vehicle capacity and cross-docking into account. Unilever runs this from annual DRP through daily replenishment and transportation plans, with manual intervention only for exceptions.
Because each store target already accounts for what the DC can supply, planners rarely need to adjust it: over 95% of Solvoyo’s plans execute untouched. To see a multi-echelon model built on your own network, request a demo.
Frequently Asked Questions
Multi-echelon inventory optimization (MEIO) sets inventory targets for all tiers of a supply network at the same time, such as plants, central DCs, regional DCs and stores. It decides how much buffer stock to hold and at which tier, so the network meets its end-customer service targets with the least total inventory investment.
Single-echelon optimization sets each location’s stock on its own, assuming its supplier always has stock. Multi-echelon optimization plans all locations together and accounts for how service at an upstream location changes the lead time and stock-out risk downstream. That avoids duplicated buffers and puts stock where it protects the most demand.
You need the network structure and lanes, lead times and their variability by lane, demand history and forecast error at customer-facing locations, service targets by segment, holding, ordering and transport costs, constraints such as pack sizes and capacity, and current inventory positions including stock in transit.
