Cutter & Buck's Second Automated DC Needs a Two-Site Inventory Allocation Rule
Cutter & Buck has added a second automated distribution center, but the important operational change is not simply more robotic capacity. It is a new decision network. Every order and every unit of inventory can now be assigned between Grand Prairie, Texas, and Renton, Washington—and a weak rule will merely move cost from the warehouse to transportation, working capital, or customer service.
Supply Chain Dive reports that the Texas site targets 1,256 bin presentations per hour, while a simulation averaged 1,284. Its AutoStore installation has 50,502 bins, 32 robots, and 13 ports, with capacity for as many as 10 million apparel, footwear, and accessory units. Renton began with more than 34,000 bins and 54 robots in 2023, then added 9,962 bins as demand and SKU count grew.
Those numbers describe two capable nodes. They do not determine which node should hold a medium polo, decorate it, and ship it to a customer tomorrow. That requires an allocation rule built around demand, inventory, cutoff times, and constrained capacity.
Stop Treating the Second DC as Overflow
An overflow model sends work to Texas only when Washington is full, or vice versa. That wastes the geographic advantage of a two-site network and creates erratic utilization. A better design gives each SKU-location pair a preferred stocking position while preserving the ability to rebalance.
Start with four inputs:
- SKU velocity: Rank items by order lines and units, but separate blank garments from decorated products whose embroidery or finishing requirements change the work content.
- Geographic demand: Map recent demand by delivery region and season. Texas is already Cutter & Buck's top sales market, making proximity a meaningful factor rather than a theoretical benefit.
- Promise and cutoff time: Calculate whether each site can complete picking, decoration, packing, and carrier handoff before the order's service deadline.
- Usable capacity: Consider available inventory, ports, robots, labor, decoration capacity, dock appointments, and carrier capacity together. Nominal bin capacity alone is not a fulfillment promise.
The default rule should place fast-moving regional SKUs near demand, duplicate only items whose service benefit exceeds the carrying cost, and assign slow movers to one primary node with planned transfer options. Recalculate the policy on a fixed cadence and after material demand, lead-time, or capacity changes.
Score Every Order Before Routing It
Order routing should optimize the complete customer outcome, not a single warehouse metric. A practical score can combine expected delivery time, fulfillment cost, inventory risk, workload, and split-shipment exposure.
First, eliminate any site that cannot meet a hard constraint: insufficient stock, missed decoration cutoff, unavailable service, or a capacity ceiling. Then score the feasible options. Favor the nearest node only when it can protect the customer promise without consuming stock needed for higher-confidence regional demand.
Multi-line orders deserve special treatment. Sending two packages may make both warehouses look productive while increasing parcel cost and giving the customer two delivery experiences. Apply a consolidation test: compare the cost and promised arrival of one complete shipment against a split. If waiting briefly for replenishment preserves the promise, consolidation can win. If one delayed line would make the entire order late, split only with an explicit cost and service rationale.
The rule also needs a confidence level. Demand history, inventory accuracy, inbound reliability, and processing-time variability should determine how aggressively the system routes. Low-confidence orders should enter a planner queue instead of receiving an automatic assignment that looks precise but rests on weak data.
Measure the Network, Not Two Warehouses
Local throughput can hide network failure. Track these measures across both facilities:
- Split-shipment rate and cost: Count orders fulfilled from both nodes, incremental packages, and added freight expense.
- Inventory duplication: Measure units and working capital stocked in both locations, segmented by velocity and margin.
- Promise accuracy: Compare the date shown at order capture with actual delivery, including orders rerouted after release.
- Automation utilization: Monitor robot, port, and downstream station constraints by interval—not merely daily averages.
- Rebalance activity: Record transfers, emergency replenishments, aged inventory moved, and stockouts avoided.
External conditions matter too. Logistics Management notes that inland markets represented roughly 80% of U.S. industrial net absorption year to date, while trade with Mexico has strengthened Texas distribution corridors. National industrial vacancy fell to 6.8% in the third quarter. A two-node policy therefore should account for the value of already-secured capacity and avoid assuming that emergency space can always be added cheaply.
Build an Inventory-Imbalance Exception Workflow
A TMS/WMS workflow should detect imbalance before it becomes a customer problem. Trigger an exception when projected days of supply at one node falls below its threshold while the other holds transferable surplus, or when demand changes enough to invalidate the current stocking policy.
The alert should contain the affected SKUs, forecast horizon, on-hand and committed stock, inbound orders, projected stockout date, transfer lead time, transport cost, and customer promises at risk. The system can then recommend one of four actions: transfer inventory, redirect an inbound purchase order, temporarily change order-routing preference, or accept the imbalance and protect stock for priority demand.
Ownership matters. Inventory planning should approve the quantity; warehouse operations should confirm handling and receiving capacity; transportation should select the transfer mode and delivery window; and customer service should receive notice only when promises may change. Every override needs a reason code and expiration time so a temporary response does not quietly become permanent policy.
Turn Automation Data Into Allocation Learning
Cutter & Buck used years of operating experience in Renton to shape the Texas design rather than copying the first installation. The same learning loop should govern allocation. Review forecast error, actual task times, carrier performance, and override outcomes every week. Use the results to update processing capacity, safety stock, cutoffs, and routing weights.
The payoff is not maximum activity at both sites. It is a network that places the right inventory near likely demand, absorbs regional variation, limits duplicate stock, and keeps customer promises at the lowest defensible total cost.
Ready to coordinate inventory, warehouse exceptions, and transportation decisions across a multi-DC network? Request a CXTMS demo to see how shared rules and real-time execution data can turn two facilities into one operating system.
