Shipsy’s linehaul optimization engine designs the inter-hub trip plan — which shipments consolidate onto which trip, on which lane, on which vehicle type, leaving at which cut-off, with which routing between hubs. It does this continuously, not just at daily plan time: as shipments accumulate at origin hubs, the engine reoptimizes trip plans so each departing vehicle is as full as it should be without breaching the next hub’s cut-off. Enterprises deploying this engine typically see 15–25% line-haul cost reduction and substantially better SLA on inter-hub transfers.

Why we built this

Middle-mile is where network economics are made or lost. A parcel sitting at a hub for 18 hours because it missed the cut-off of a half-full trip has ruined the unit economics of that shipment even if every other leg runs perfectly. And line-haul planning is done manually in most enterprises — typically by a senior planner running a spreadsheet of known lanes, asking dispatchers “when’s the truck leaving Hub A for Hub B”, and rubber-stamping trips that are either under-loaded or over-committed.

The mechanism we built replaces that spreadsheet. It treats line-haul as a continuous consolidation problem — which shipments, onto which trip, onto which lane, against which cut-off — and solves it with operational-research horsepower plus live signals.

How it works

The engine operates across four layers:

Layer 1 — Lane and trip master. Every inter-hub pairing in the network is modeled as a lane, with base transit times, supported vehicle types, cost structures (own fleet, contract, spot), and SLA commitments. Trips are instances on lanes — a trip is “Vehicle X departing Hub A at 10pm for Hub B, with a cut-off of 9pm for shipments.”

Layer 2 — Shipment-to-trip assignment. As shipments accumulate at a hub, the engine assigns each to a candidate outbound trip. Assignments use a multi-objective cost function: SLA fit (will this trip arrive before the shipment’s service-level deadline at the next leg?), vehicle fill rate (can this shipment help convert a half-full trip to full?), vehicle type fit (weight, volume, SKU compatibility, hazmat), and cost (own vehicle vs spot hire). The engine is continuously re-evaluating — a shipment initially assigned to Trip A at 10pm can be reassigned to Trip B at 9pm if Trip B suddenly has capacity and A is running under-loaded.

Layer 3 — Dynamic consolidation and trip creation. The engine watches trip fill rates against cut-offs. Three actions are available: accelerate (advance cut-off if the trip is full early), consolidate (merge two under-loaded trips that share a destination if SLA permits), or create (spin up an additional trip if the incoming volume exceeds planned capacity). Each action is proposed with explicit trade-offs (cost saved, SLA impact, vehicle utilization) and either auto-executed within thresholds or reviewed by a planner.

Layer 4 — Multi-leg routing. For shipments that need to traverse multiple hubs (origin hub to transit hub to destination hub), the engine plans the full path — not just the next leg. It optimizes the combined cost and timing of multi-leg paths, considering transit-hub dwell times, cross-dock windows, and alternate routings that trade cost for speed or vice versa.

All trip plans feed hub operations (inbound wave planning, outbound readiness) and feed carrier performance scorecards for multi-carrier line-haul networks.

Here’s the network flow at a glance:

graph LR O1[Origin hub A] --> C[Consolidation point] O2[Origin hub B] --> C C --> T{Trip assembly} T --> L1[Linehaul lane 1] T --> L2[Linehaul lane 2] L1 --> D1[Destination hub X] L2 --> D2[Destination hub Y]

Early results

Enterprises deploying linehaul optimization typically report, within 90 days:

A leading Western European parcel operator with 50%+ national market share uses this engine across its domestic inter-hub network to consolidate 8M+ shipments through a tight hub-and-spoke design.

What’s next

Three upgrades: carrier marketplace integration for spot-hire decisions on under-served lanes, trip-level emissions accounting so cost functions can trade emissions against cost within policy envelopes, and predictive capacity for peak season — the engine forecasting lane-level capacity crunches a week ahead and proposing pre-booked contracted capacity.