Here's the conclusion up front: in 3PL multi warehouse inventory allocation, only one rule survives contact with reality — allocate by outbound demand, not by splitting inventory evenly. The warehouse closer to the orders, the one that ships faster, holds more stock. An even split looks fair. In practice both warehouses end up paying for each other's stockouts.

I've seen this exact mistake three times. The third was in spring 2025, with a home-goods client running two Southern California warehouses — one in Ontario, east of Los Angeles, and one 120 kilometers away in Riverside. The client insisted on half and half, "to spread the risk." Peak season arrived. The bestseller went out of stock in Ontario while the same SKU piled up to the ceiling in Riverside. Transfer trucks ran between the two buildings every day, and the client asked why the transfer bill was bigger than the storage bill.

Why an Even Split Always Breaks

3PL multi warehouse inventory allocation looks like a math problem. It is actually an organization problem.

Two warehouses' stock is not two numbers. Behind them are two crews, two equipment sets, two shipping rhythms. The Ontario building sits near the port, so inbound is fast, but space is tight. Riverside is big and cheap, but farther from most delivery addresses. Splitting stock fifty-fifty pretends the two buildings are identical. They are not, not for a single day.

Order geography is never even either. About 70 percent of this client's orders went to West Coast metro areas, where Ontario had the edge. The remaining 30 percent headed inland, where Riverside was the better ship point. The even split left Ontario short and waiting on transfers every day, while Riverside sat on stock nobody was ordering. Both buildings looked underutilized on paper while total inventory looked "sufficient."

The rookie mistake is allocating by SKU count instead of by sell-through speed. SKU A moves 5,000 units a month. SKU B moves 200. Give each half the stock and A is gone in three days while B sits for six months. The unit of allocation is outbound demand, not SKU count.

Writing 3PL Multi Warehouse Inventory Allocation Rules Into the System

Allocation rules need to be executable by the system, explainable by customer service, and signable by the client. Our standard approach is four layers, checked top to bottom.

Layer one is ship-from proximity. Take the last 90 days of order ship-to ZIP codes and assign each SKU a primary warehouse. The primary holds roughly 70 percent of the stock; the secondary holds about 30 percent as backup and overflow. Primary status is not permanent — recalculate quarterly from order distribution.

Layer two is ABC classification. A-items, the fast movers, are stocked in both buildings, heavier in the primary. B-items live only in the primary; if an order needs one from the secondary region, transfer it. C-items, the slow movers, sit in one building only, so they don't eat rack space in two places.

Layer three is separate safety stock. A common client error is setting two buildings' safety stock at the single-warehouse quantity. Each building faces stockout risk independently, so safety stock has to be calculated per building from its own demand variability and replenishment lead time. Total safety stock across two buildings is genuinely higher than for one. That's the real cost of multi-warehouse, and hiding it helps nobody.

Layer four is the exception list. Promotions, new products, seasonal items, clearance — none of them follow the standard rules. Each gets its own allocation plan. Promotional stock is decided in a meeting a week before launch, noted in the system, and the rule is retired the day the promotion ends.

My recommendation: keep the rules in a configuration table in the WMS, not hard-coded. Each row reads: SKU, primary warehouse, secondary warehouse, primary share, ABC class, exception reason. When the client wants a change, edit the table and leave a change record — who, when, why.

SKU type Primary stock Secondary stock Transfer trigger
A — fast mover ~70% ~30% Primary below 5 days of sales
B — medium All None Order ships to secondary region
C — slow mover All in one building None No transfer as a rule
Promo / new item Per campaign plan Per campaign plan One-time pre-positioning

When 3PL Multi Warehouse Transfers Trigger, and Who Pays

Transfers are where the arguments start. Clients see transfers as the 3PL's internal business and resist paying. The 3PL sees transfers as the consequence of the client's allocation strategy and expects someone to cover the cost. In my experience, if it isn't written in the contract, the argument is only a matter of time.

Write the triggers as hard rules — no "at our discretion." Our three standard triggers: a SKU's available stock in the primary falls below five days of projected sales; a SKU in the secondary has zero outbound movement for 30 straight days; or the client requests a bulk move in writing. Once triggered, execute. Don't wait for approval — the stockout losses during two days of waiting cost more than the transfer itself.

Cost sharing goes in a contract appendix, in three tiers. Tier one: transfers caused by a 3PL allocation error — wrong warehouse assigned in the system, receiving put away in the wrong building — the 3PL pays. Tier two: transfers caused by a client strategy change — a different primary warehouse, a revised promotion plan — the client pays. Tier three: routine replenishment transfers, billed at actual cost with the unit price in the rate card, and a monthly transfer statement attached: date, SKU, quantity, origin, destination, amount.

A hypothetical worked example shows why the transfer bill stings. Assume one transfer of 200 units over 120 kilometers, with trucking, loading labor, and system handling totaling 380 dollars — an illustrative figure, use your real quoted rates. Twenty transfers a month is 7,600 dollars. Over a year that's more than 90,000 dollars. Put that number in front of a client and they immediately understand: the cost of a wrong allocation rule is far bigger than the transfer fees themselves.

The system accounting matters too. A transfer in the WMS is not just "moving stock." One transfer order has to do three things at once — outbound from the origin, in-transit status, and inbound at the destination. In-transit stock must not be promised to orders. That's the most common system failure I've seen. In one project, in-transit stock got promised by both buildings at once. Oversold orders. Customer service spent three days apologizing.

Go-Live Night Checklist

As always, the go-live night checklist — check each box:

  1. The allocation configuration table is signed by the client, with the exception list attached separately
  2. Primary and secondary definitions are in the system; changing a primary requires a change request, not a casual edit
  3. Transfer thresholds are set, and the 5-day sales alert and 30-day zero-movement alert have both been tested
  4. Transfer cost-sharing terms are in the contract, with unit prices in the rate card
  5. In-transit locking is tested: stock in transfer cannot be promised or picked
  6. The client can see inventory levels per warehouse, not just a combined total

One counterintuitive point to close on: fewer transfers doesn't mean better allocation. Two warehouses with zero transfers in a month are probably both hoarding safety stock, with total inventory far higher than it should be. The healthy metric isn't zero transfers. It's transfers that happen inside the rules, with costs accounted for and the client in agreement. The goal of allocation isn't to never transfer. It's to make every transfer traceable.

By Lin Xiao, WMS Implementation Consultant