When an inbound flight runs late, every onward connection behind it is suddenly at risk. This demo shows how a multi-hub airline group can re-accommodate all affected passengers in a single mathematical pass — using a Python backend with Google OR-Tools' min-cost flow solver, the same class of optimisation that runs in real operations control centres. All data is synthetic; nothing here touches live systems.
Pick one of the pre-built delay scenarios below. The backend runs the moment you click.
The engine flags which passenger groups will miss their connection — and how likely a miss really is.
Read the rerouting proposals, then move the risk slider and watch the recommendation change live.
Each scenario delays one inbound feeder flight to a hub. New to the demo? Start with Scenario B — it is the clearest demonstration of the solver at work: four passenger groups stranded at Frankfurt, all competing for the same seats out.
Connecting to the OR-Tools backend and loading scenarios…
python engine.py. In production, point this page at the deployed API with
?api=https://your-rerouting-api.example.com.
Compared the revised arrival against every onward departure — minimum connection time plus the hub's MPR miss-probability curve.
Turned today's remaining schedule into a graph where every node is an (airport, time) pair and every flight is an arc with seat capacity.
OR-Tools min-cost flow assigned all groups at once, never overselling a seat — ….
Ranked each itinerary by delay + stops + misconnection risk and flagged anything too poor for automation.
A connection is not simply "made" or "missed". Below the hard minimum connection time it is physically impossible; above it, the engine uses the misconnection prediction rate (MPR) — the historical probability of missing the flight given that much transfer time at that specific hub. Gate distance, passport control and terminal changes are all baked into the curve.
Run a scenario above — the disruption details appear here.
Each group of passengers on the delayed flight, with its onward connection and risk verdict.
One card per passenger group: the broken itinerary, the engine's proposal, and the score that justifies it — lower is better. Click any card to see the alternative paths the engine weighed against each other, including exactly how much misconnection risk contributed to each score.
The engine's proposals appear here after you run a scenario.