Predicting which airfreight route moves next — and why.
Every route gets a signal.
The problem
Benchmarks, schedules, live flight signals, macro indicators and disruption events sit in separate systems. Decisions are still made from lagging views.
A forward market needs a shared evidence layer — one model of the system that everyone can interrogate.
Today's fragmented views
Benchmarks
What cleared?
Flight schedules
What capacity is planned?
Live flight signals
What is actually moving?
Macro & disruption signals
What pressure is building?
Reports
What happened?
A shared forward view
Which route is changing next — and why?
How it works
Airfreight is aircraft, hubs, lanes, cargo and rates in constant motion. GraphCargo reads them as a single connected system and turns that reading into route-level foresight.
Signals → Evidence graph → Forecast corridors → Weekly self-check
Macro flows, trade indicators, capacity movements and disruption events are read together, in one frame. Each series carries a record of when it became known — the goal is to treat every value by its documented availability; 167 series are currently being re-measured under a conservative time contract.
Every relationship the system relies on is stored as evidence — which signal, which run, which control it survived. Any forecast can be traced back to why it says what it says.
Routes don't get a single magic number. They get a corridor — the range a rate is expected to stay inside — together with the drivers behind the move.
Every week, live forecasts are scored against what the market actually did. The scorecard is part of the product — not a drawer we open when it flatters us.
Many methods, one graph
Independent methods look at the same lane from different angles — and write their results as evidence into one graph. A claim only gets strong where methods converge; every reading stays traceable down to the single signal.
Estimates the range a rate should stay inside — a band, not a single number.
Read whether a lane is more likely to firm or soften over the coming weeks.
Independent reference models that must be beaten, not decorated around.
Every result has to beat shuffled versions of itself before it may count.
Today this graph makes every statement traceable — which signal, which run, which control. The market-structure layer, the projection neighbourhood learning grows from, is our second projection — it is not built yet.
The verification chain
Walk the path every figure has to survive — from source to use. Click through the stations; each one exists to kill a specific way of fooling ourselves.
Station 1 of 7
Every source carries what it is: licensed official full history for five routes, a publicly read ticker, 127 signal series, real shipments, flown capacity.
Worked through
The official rate history and the scraped ticker never blur into one series — status and licence stay visible on every value.
Only after surviving this chain is a number released — and what fails re-testing is retracted, publicly.
Evidence discipline
Forecasting airfreight is easy to claim and hard to prove. So the lab holds itself to research discipline — and the discipline, not the highlight reel, is what we show.
Pre-registered
Evaluation criteria are written down before a model runs. Success is defined in advance — not discovered afterwards.
Walk-forward
Models are scored only on data that lies strictly after everything they were trained on. No hindsight, no leakage.
Randomized controls
Every signal has to beat shuffled and permuted versions of itself before it counts as real.
Retraction
Findings that fail re-testing are withdrawn and recorded as withdrawn. What didn't survive is part of the evidence.
A number is released — into the product, and onto this page — only after it survives scrutiny.
Everything shown runs on licensed trial data from selected partners — the official TAC full history for five routes, and real shipment data. The proof runs on real market data; production breadth arrives with the full feeds after closing.
Who it's for
A forward market view of the lanes you sell — not last week's rate sheet. See which corridors are tightening and what is driving the move, as shared ground for customer and route conversations.
Read pressure across the network as it forms — route by route, with the evidence behind every reading in plain sight.
One shared forward picture of market, capacity and evidence — instead of a stack of systems that disagree about yesterday.
GraphCargo provides a shared, evidence-backed market picture — it does not issue pricing, quoting or trading recommendations.
Where this goes
What we are building next — stated as plan, held to the same verification chain before any of it may speak with numbers.
Strikes, closures, geopolitics: an LLM — a large language model, reading text at scale — places each event next to the recorded impact of comparable past events. An assistive reading with every source on record, feeding the same evidence graph — not a shortcut past the controls.
Forward-looking schedule capacity from OAG — the published flight plans of the airlines — joined with flown actuals from ADS-B, the live transponder record of what really operated. Planned supply and real supply, side by side per lane.
Together, these let us work the supply curve out of the network picture — planned capacity and flown capacity, read side by side per lane. Stated as intent: it enters the product only after it survives the verification chain.
Not just a dashboard
GraphCargo integrates through MCP — the open protocol that lets AI assistants and internal systems query live sources directly. Corridors, drivers and evidence arrive where your analysts and systems already work. The current market picture for a lane is a query, not a meeting.
The ambition
The Bloomberg terminal for airfreight — explainable down to the single signal.
Stated as ambition, not as a claim — the verification chain decides what we get to say.
Request a briefing
A briefing is a working session, not a demo reel: the system, the evidence discipline, and what a shared forward view would mean on the routes you care about.
Direct contact
Partners get access to the live product surface and the evidence record behind it.
Partner access