How Shipsy plans vehicle loads with 3D packing and live capacity utilization

Shipsy’s load-planning engine packs vehicles in three dimensions — not by flat weight, not by nominal pallet count, but by actual cubic feasibility with stacking rules, orientation constraints, axle-weight limits, and drop-sequence preservation. The result: 10-25% higher fill rates and zero “we can’t fit it on the truck” surprises at the dock.

Why we built this

Two legacy patterns dominate load planning, and both leak money. The first is “weight-only” — a 12-ton vehicle gets 12 tons of cargo regardless of whether it cubed out at 8 tons or had 40% airspace left. The second is “pallet-count” — 26 pallet positions per trailer, done. Neither accounts for irregular freight, mixed SKUs, fragile-on-top stacking rules, or the need to deliver stops in route order (last-loaded must be first-unloaded).

Large-format retailers, fast-moving beverage distributors, and big-and-bulky networks lose margin every day to suboptimal loads. A major beverage bottling group operating primary + secondary distribution and a global big-and-bulky retailer leading in furniture and home goods both pointed to fill-rate gaps as their largest unaddressed freight-cost lever.

How it works

Cargo profile intake. Every SKU in Shipsy’s load planner carries a cargo profile: dimensions (L x W x H), weight, stackability (can it bear load on top? how much? how many tiers?), orientation (can it be tipped? which side up?), fragility, hazardous-material class, temperature zone, and drop-sequence priority. Master data is typically imported from ERP or WMS on first onboarding and maintained via a structured change feed.

Vehicle profile intake. Each vehicle in the fleet has a profile: internal dimensions (length, width, height), max gross weight, axle-weight limits (front, middle, rear), door access pattern (rear-only, side-load, tailgate), temperature zones if reefer, and any special equipment (liftgate, tail-lift, strapping kit).

3D bin-packing with constraints. The core is a 3D bin-packing solver that respects all the above constraints simultaneously. It’s not a greedy packer — it iterates across hundreds of candidate configurations and selects the one that maximizes fill rate while respecting stacking rules, axle balance, and drop sequence. Classic greedy packers hit ~75% fill; Shipsy’s constrained solver regularly hits 88-94% on real enterprise freight.

Drop-sequence preservation. For multi-drop routes, cargo is loaded in reverse drop order — the last stop goes in first, the first stop comes out first. The solver treats drop sequence as a hard constraint: it will find a lower-fill-rate plan rather than load a 4th-stop pallet behind a 1st-stop pallet. Drivers never have to climb over freight.

Axle-weight balance. For heavier trucks and cross-border freight where axle-weight fines are material, the solver balances weight distribution front-to-rear and side-to-side. Output includes the predicted axle-weight readout so the dock can sanity-check against scale.

Integration with Astra. Load plans feed into Astra, Shipsy’s planning agent, which combines load plans with routes. When Astra re-optimizes a route mid-execution (see dynamic rerouting during execution), it respects the load plan — it won’t reorder drops in a way that makes the load physically undeliverable.

Cross-dock and consolidation awareness. For networks running cross-dock operations (see cross-dock orchestration), the load planner handles the inbound-to-outbound re-consolidation: freight arriving on an inbound linehaul gets re-packed into outbound micro-consolidation vehicles using the same constrained solver, with drop-sequence awareness on the outbound leg.

Dock-side visualization. Dock workers get a visual load plan — a 3D rendering of the trailer with each SKU placed in its intended position, numbered in load order. The visual reduces the “I think it goes here” guesswork that destroys fill rate even when the plan is optimal.

Exception handling. If actual cargo varies from plan (SKU shorter-ship, damage, unexpected return-to-dock), the solver re-runs in seconds with the actual freight and proposes an adjusted plan. No dock clerk re-does math on a clipboard.

Early results

Operators moving from weight-or-pallet-based planning to 3D constrained planning typically see: fill rate up 10-25 percentage points (translating to 10-20% fewer trucks per shipment volume), driver complaint rate (about “impossible loads” or “wrong sequence”) down sharply, and fewer axle-weight fines on cross-border lanes. A global big-and-bulky retailer leading in furniture and home goods has used this mechanism as a core lever in improving cost-per-stop on last-mile bulky delivery.

For beverage and FMCG networks, mixed-SKU loads (different pack sizes, different brand families) hit fill-rate targets that flat-rate planners can’t approach. A major beverage bottling group reports fill-rate improvements feeding directly into their depot-to-retail cost structure.

What’s next

Next release adds returnable-packaging (RPM) awareness — kegs, crates, and roll cages on their return journey are planned as part of the same load optimization, reducing the “empty backhaul” problem that plagues RPM-heavy networks. See returnable packaging tracking (RPM).