U.S. wholesalers are running lean. The merchant wholesalers inventories-to-sales ratio was 1.18 in August 2026, seasonally adjusted, against 1.28 a year earlier (U.S. Census Bureau via FRED, WHLSLRIRSA). In May 2026 it touched 1.15, a level last seen in April 2012 (U.S. Census Bureau via FRED).
Now restocking is starting, and where that inventory lands matters as much as how much of it there is.
Reading the Ratio
The ratio divides wholesalers' end-of-month inventories by their monthly sales. A reading of 1.18 means stock on hand equal to roughly 1.18 months of sales across U.S. merchant wholesalers (U.S. Census Bureau via FRED). It is an aggregate across every wholesale industry, so it describes the environment you operate in, not your own position.
The path to today is worth seeing. The ratio spiked to 1.66 in April 2020 when sales collapsed, rose again to 1.40 by December 2022 as inventory built up, and has fallen most of the way since (U.S. Census Bureau via FRED):
| Month | Ratio (SA) | Context |
|---|---|---|
| Apr 2020 | 1.66 | Series high since 1992; sales collapse (FRED) |
| Dec 2022 | 1.40 | Post-pandemic inventory build (FRED) |
| Aug 2025 | 1.28 | One year ago (FRED) |
| May 2026 | 1.15 | Lowest since April 2012 (FRED) |
| Aug 2026 | 1.18 | Latest; next release November 17, 2026 (FRED) |
Freight data points the same way. Cass reports freight shipments up 1.9 percent year over year in September 2026, the second straight annual gain, helped by "some inventory restocking as container imports surge" (Cass, September 2026). Thin buffers combined with returning volume is the classic setup for a restocking cycle. It also creates the risk of putting new stock in the wrong places.
In a restocking cycle, the question is not only how much inventory to buy. It is which node in the network should hold it.
Why Node-by-Node Safety Stock Falls Short
Most companies set safety stock one location at a time. Each distribution center, warehouse or plant gets a target based on its own demand variability, its own replenishment lead time and its own service goal. This is single-echelon planning, and each node is optimized as if it stood alone.
The flaw was identified more than six decades ago. In their 1960 paper in Management Science, Andrew Clark and Herbert Scarf noted that inventory models had been "almost without exception" devoted to a single installation, and that this misses a basic fact of multi-site systems: when one site orders from another, delivery time depends not only on the transit time but "also by the availability of stock at the second installation" (Clark & Scarf, "Optimal Policies for a Multi-Echelon Inventory Problem," Management Science, 1960). In other words, every node's effective lead time depends on stock upstream. Plan nodes separately, and each one buffers against the others. Inventory gets duplicated across the network while service can still fail.
What Multi-Echelon Inventory Optimization Does
Multi-echelon inventory optimization (MEIO) plans the whole network at once. An echelon is a tier: suppliers, plants, central distribution centers, regional warehouses, forward stocking locations. MEIO treats inventory at every tier as one connected system and decides together how much to hold and where.
The foundational practical model comes from Stephen Graves and Sean Willems at MIT. Their 2000 paper starts from the problem every operations leader recognizes: "in complex supply chains, it is not always obvious where to hold safety stock to minimize inventory costs and provide a high level of service to the final customer" (Graves & Willems, "Optimizing Strategic Safety Stock Placement in Supply Chains," M&SOM, 2000). They model the supply chain as a network in which each stage commits a guaranteed service time to its customers, and they optimize where strategic safety stock should sit (Graves & Willems, 2000).
| Single-echelon (node by node) | Multi-echelon (MEIO) | |
|---|---|---|
| Unit of planning | Each location on its own | The full network as one system |
| Lead time | Treated as given by the supplier | Depends on upstream stock and service commitments (Clark & Scarf) (Graves & Willems) |
| Main question | How much at this node? | Which nodes should hold buffer, and how much at each? (Graves & Willems) |
| Typical failure | Duplicated buffers across tiers | Needs network data: lead times, demand variability, costs and service times |
The method has been applied in industry. Graves and Willems describe use by product flow teams at Eastman Kodak to reduce finished goods inventory, target cycle time reduction efforts and determine component inventories (Graves & Willems, 2000). We do not cite a general percentage benefit for MEIO here: results depend on the network, and published figures vary by case.
When MEIO Pays Off
MEIO is not needed everywhere. Based on the structure of the models above, it earns its complexity when several conditions hold together:
- Several tiers of stocking locations. The modeling problem only exists when stages feed other stages (Clark & Scarf).
- Long or variable lead times between tiers. These drive the interaction between upstream availability and downstream service that single-node models ignore (Clark & Scarf).
- Demand or forecast uncertainty that varies by product and location. Graves and Willems model safety stock precisely as protection against this uncertainty (Graves & Willems, 2000).
- Clear service commitments to the end customer. The model is built around guaranteed service times between stages (Graves & Willems, 2000).
A single warehouse serving one market gains little. A national network with central and regional distribution centers, long inbound lead times and many SKUs is where the method is designed to work.
Connecting MEIO to the Restocking Cycle
A restocking cycle is the moment network-level placement decisions are made, whether deliberately or not. Every replenishment order decides which echelon absorbs the new stock. Three practical implications follow.
1. Decide the echelon before the quantity. Stock held at a central node can serve several downstream locations; stock held forward is closer to the customer but committed to one region. Choosing where strategic safety stock sits across the network is the problem the Graves and Willems model solves (Graves & Willems, 2000). Pushing everything forward by default risks the wrong stock in the wrong region.
2. Set service times, not just stock targets. In the Graves and Willems framework, the service time each stage promises the next determines where safety stock is needed (Graves & Willems, 2000). Agree these internally before the restocking volume arrives.
3. Plan the freight with the inventory. Restocking moves freight: Cass notes that the normal seasonal pattern would put its shipments index up 4 percent year over year in October (Cass, September 2026). Placement decisions set how many miles that freight travels and how much of it later has to be rebalanced.
Questions Every Distribution Leader Should Be Able to Answer
- How is your inventories-to-sales ratio moving against the wholesale benchmark? The aggregate was 1.18 in August (FRED).
- Is safety stock set node by node, or across the network?
- Do your regional sites and central DCs both buffer for the same demand uncertainty? If so, you are holding the same buffer twice.
- What service time does each tier commit to the next? If no one can answer, safety stock is being set without its key input (Graves & Willems).
- Which echelon will absorb this quarter's replenishment orders, and who decided?
The Bottom Line
Wholesale inventory is lean by historical standards and freight volume is recovering. Restocking is the point where companies either rebuild inventory where it serves the whole network or rebuild the same duplicated buffers they had before. The theory for planning it as one system has existed since 1960; the practical model since 2000. What most networks lack is the decision to use it.
Place inventory as a network. Not as a list of warehouses.
