Multi-Warehouse Sourcing: How an OMS Decides Which Warehouse Ships

Bottom line up front: When an order lands, an OMS runs it through exactly four gates, in this order — inventory availability, transit time and cost, warehouse capacity, and special rules. Any gate can veto a warehouse. If the inventory gate says no, the warehouse is out; it never even gets judged on distance. Get this order right and sourcing mostly takes care of itself. Get it backwards and freight bills and oversells will take turns ruining your month.

I've seen three clients treat sourcing as "ship from the closest warehouse." All three paid tuition.

Here's one from November 2024. A small-appliance seller in California, three warehouses: Los Angeles, Dallas, New Jersey. Their OMS had exactly one sourcing rule: pick the warehouse nearest the ship-to address. Sounds perfectly reasonable — shorter distance, cheaper freight, faster delivery, right?

Black Friday morning proved otherwise. The LA warehouse was closest to most buyers, but the hot SKUs were sitting in Dallas; LA had only a few dozen units. The OMS dutifully routed about 80% of orders to LA. By mid-afternoon LA was buried: a six-hour picking backlog, UPS trailers sitting in the yard waiting for loads that never got staged. Meanwhile Dallas sat on 20,000 units nobody touched. Day two was worse — LA hit zero, but the OMS was still allocating off yesterday's inventory snapshot. Orders kept flowing in, the warehouse kept saying "no stock," and customer service drowned in delay apologies.

We got called in early December to clean it up. One look at the logs told the story: not an OMS problem, a sourcing-rule problem. The fix was straightforward: promote inventory availability to the first gate, add a capacity circuit breaker, and move inventory sync from once-daily to real-time reservation. Two weeks to deploy, and the Christmas season held.

That project is why I keep saying: "closest" is the fourth consideration in sourcing, and three gates stand in front of it. Here they are.

The Four Sourcing Gates

The order is non-negotiable. Availability is always gate one: a warehouse with no stock is disqualified outright and never enters round two. Putting "nearest" first is exactly the mistake that California seller made.

Gate two is where time and cost fight it out, and here's the counterintuitive part: the closest warehouse is not always the cheapest one to ship from. Dallas to Chicago via UPS Ground is a two-day lane; New Jersey is geographically nearer but can land in a worse zone-plus-surcharge combination. Or the LA site is a 3PL charging per-order handling fees while Dallas is your own building with none. A real sourcing engine ranks gate two by total fulfillment cost, not straight-line distance. In one project, distance-only sourcing ran about $180K a month in freight; switching to total-cost ranking brought it under $130K. That $50K gap was pure "closest warehouse" tax.

Configuring Sourcing Rules: The Go-Live Night Checklist

This is the checklist I walk through the night before every multi-warehouse go-live. In order, no skipping:

1. Lock down your inventory definition first. The inventory the OMS sees must be available inventory — on-hand minus allocated, minus reserved, minus QA hold. Virtual inventory (inbound POs, goods in transit) must never enter the sourcing pool, or if it does, it goes into a separate pre-sell pool that never mixes with sellable stock. I've seen this exact failure three times: virtual stock counted as available, OMS allocated it, the warehouse said "we don't have it," and CS ate the apologies. All three.

2. Write the rule engine in gate order. Build the rule group in the OMS admin: (1) inventory filter — drop warehouses with no stock; (2) cost-and-time ranking — sort survivors by total fulfillment cost, low to high; (3) capacity check — downweight or exclude warehouses past their daily throughput cap; (4) special-rule overrides — VIP customers pinned to a site, hazmat only from certified sites, cross-border orders from the bonded warehouse. Every rule gets a clear "if this, then that." No "preferably," no "where possible."

3. Configure split-shipment rules separately. One order, three line items: warehouse A has two, warehouse B has one. Split it or hold the whole order? Splitting is fast but doubles the shipping cost; holding protects cost but can blow the delivery promise. My default: allow splits for high-value or time-sensitive orders (cap it — two parcels max), and for low-value orders ship complete from the warehouse holding the most, backordering the missing line. If you don't decide this before go-live, peak season CS will be answering "why did my order arrive in two boxes" a thousand times a day.

4. Add circuit breakers and a fallback. Set a daily processing cap per warehouse (historical peak times 1.2); when a site hits it, new orders route to the next-best site automatically. Then add a catch-all: any order that clears all four gates with no warehouse assigned drops into a manual review queue with a 15-minute SLA. An OMS without a fallback turns every warehouse into a single point of failure.

5. Load-test before launch. Replay three days of your last peak season's order volume through sourcing in a test environment. Watch three numbers: oversell rate (must be zero), average fulfillment cost per order, and peak load on your busiest warehouse (keep it under 85%). Sign off only when all three pass.

The Other Story: When Sourcing Is Right, the Savings Are Real Money

March 2025. A pet-supplies retailer in Texas, four warehouses: Dallas, Atlanta, Chicago, Los Angeles. Their sourcing was manual — a CS lead eyeballed the inventory spreadsheet every morning and assigned orders by hand. At 4,000 orders a day in peak season, she told me she dreamed about sourcing.

The setup we deployed wasn't exotic: availability first, cost second, capacity breaker third, plus one special rule — Florida orders prioritized Atlanta (hurricane-season logistics experience talking). First month live, average fulfillment cost per order dropped $1.10. At 4,000 orders a day, that's roughly $130K a month. The owner's exact words: "This rule set replaced three CS reps."

Honest caveat, though: sourcing logic is not set-and-forget. Carrier rates change yearly, capacity shifts with the season, and every new warehouse means re-ranking the gates. I tell clients to review sourcing rules quarterly, same cadence as cycle counts. Non-negotiable.

Common Failure Points: Read Once More Before Go-Live

Failure 1: Sourcing by distance only, ignoring inventory. That's story number one. Distance is a cost input, not a qualification. A warehouse with no stock is useless at zero miles.

Failure 2: Virtual inventory leaking into the sourcing pool. In-transit stock, unreceived POs, QA holds — none of it is allocatable. I've seen this pit three times. Oversell, every time.

Failure 3: No capacity circuit breaker. During peak, every order floods the "optimal" warehouse, which becomes the most miserable warehouse. A daily throughput cap is the fuse in your sourcing panel. Don't skip it.

Failure 4: Missing split-shipment rules. Either splits are forbidden and delivery promises collapse, or splits are free-for-all and freight doubles. Splits need an explicit cost threshold.

Failure 5: Rules configured, never reviewed. Sourcing rules are alive — rates, capacity, and the network all move. Configure it and forget it, and in six months it's wrong.

Sourcing comes down to four gates and a five-point checklist. The verdict was in the first paragraph, the how-to is all above, and the pitfalls are flagged. What remains is walking the checklist on go-live night, item by item. No shortcuts.

By Lin Xiao, WMS Implementation Consultant