Omnichannel fulfillment only works when one orchestration brain decides where every order ships from — not when store systems, warehouse systems, and carrier systems each make local decisions. Retailers running Shipsy’s omnichannel orchestration layer are collapsing cost-to-serve by 12-18% simply by removing conflicting routing logic across DC, store, and dark-store nodes.
The finding: siloed fulfillment is a tax on every order
Most retailers still route online orders through the nearest DC by default, even when the SKU sits in a store 3 km from the customer. The DC ships it. The store sits on dead inventory. The customer waits two days when they could have had it the same afternoon. Multiply across a year and the opportunity cost is visible on the P&L.
The retailers winning omnichannel — a MENA retail conglomerate with 80+ years operating multi-brand sports, health, and lifestyle portfolios, and a global big-and-bulky retailer leading in furniture and home goods — have moved to a single orchestration brain that evaluates every order against every node in real time. Shipsy’s Astra planning agent makes the sourcing decision against a cost-to-serve model that includes inventory carrying cost, markdown risk, fulfillment labor cost, and carrier shipping cost — not just proximity.
Why the old model breaks
Legacy OMS and TMS stacks were designed for one-way flow: warehouse to customer. They did not anticipate:
- Orders that should source from store when inventory is aging
- Ship-from-store decisions that compete with in-store demand
- Dark-store inventory that blurs the line between retail and quick commerce
- BOPIS orders that need store pick + customer hold + no shipping at all
- Returns that should re-enter store inventory, not go back to DC
When these flows are bolted onto a legacy TMS, every edge case becomes a manual override. Ops teams drown in exceptions. Store managers game the system to protect their own inventory. Central merchandising loses control of markdown cadence.
What orchestration actually does
A proper orchestration brain runs four decisions per order, in milliseconds:
| Decision | Signals evaluated | Mechanism |
|---|---|---|
| Source node selection | Store inventory, DC inventory, aging, markdown risk, picking capacity | Astra cost-to-serve model |
| Delivery mode | Standard, express, same-day, BOPIS, scheduled slot | SLA + margin optimizer |
| Carrier allocation | Service capability, carrier scorecard, cost, zone coverage | Multi-Carrier rate shopping |
| Route assignment | Driver proximity, current load, time window | Micro-cluster routing |
The first decision is the one most retailers get wrong. Shipsy’s Astra treats inventory decisions and logistics decisions as one problem, not two. A global big-and-bulky retailer routing furniture orders hit 95% first-attempt delivery rate (FADR) once the source-node decision started considering local install capacity alongside stock availability.
What to do in the next 90 days
Start with three moves. First, stand up a unified order view that ingests demand across e-commerce, marketplace, store-fulfilled, and BOPIS channels into a single order graph. Second, connect store inventory to the orchestration layer with sub-minute latency — stale inventory signals destroy fulfillment accuracy faster than anything else. Third, enable Shipsy’s Clara customer agent to proactively notify customers when the orchestration layer reroutes orders between fulfillment nodes, so CX never gets blindsided.
Retailers who skip the unified inventory view and try to optimize routing first end up with faster deliveries of wrong items. Order the sequence correctly.