The method

Engine 0.14 · run 07563456c5ce · October 2026

How the stress test reaches a break-even: what it simulates, where its numbers come from, what it leaves out, and how it is checked on your own records.

What it answers: Which spare aircraft pay for themselves over a summer, and up to what a day the rotation taken off the schedule to hold each one can earn. Beside spares of your own, it compares a standby ACMI contract and an aircraft leased for the year on the same summers. It compares plans; it does not forecast your summer.
What is simulated: Thousands of summers, day by day and base by base: 6,000 on the sample paper. Each sub-fleet loses aircraft to unplanned groundings, drawn at random: how often from its rate, how long from a spread of durations. A spare covers a grounding at its own base, or a day later at another. Without one, a base retimes a set number of lines a day; past that, flights are cancelled or leave late.

One of the 6,000 typical summers simulated, day by day: a square is an aircraft grounded for a day. On average over them all, 144 flights are cancelled with no dedicated spare, 72 with the paper’s spare.

No dedicated spare122 flights cancelled, on 47 aircraft-days

One long-range narrowbody spare at Hub44 flights cancelled, on 19 aircraft-days

How a spare is judged: Every plan meets the same simulated summers, so the gap between two plans is the spare’s, not luck. A gain or a loss is called only where the gap stands clear of the luck of the draw; short of that, the answer is no measurable difference. Calm, typical and hard seasons are read apart, never averaged; the calm and the hard ones are our picks, not percentiles.
Where the numbers come from: Your fleet: your own published reports.
What a day of flying earns: published airline accounts and ICAO’s fuel tables.
What a cancelled flight costs: its lost contribution, the compensation passengers claim where a rule owes it, such as EU 261, their care and tickets on another airline; set against EUROCONTROL’s reference costs.
Every input is sourced or marked as our assumption. None is fitted to your records: each is a figure you can read and correct, and the replay checks them against your last twelve months.
What public data cannot give: How many aircraft-days your fleet loses to unplanned groundings: no public statistic gives it. On your page it is our assumption, said as such, and the first line to correct; the replay takes it from your records.
What it leaves out: A grounding at a remote outstation: the run lets a spare reach every grounding, so it overstates what a spare avoids.
Spare engines, in place of a spare aircraft.
Crew rosters: finding a standby crew is a probability for the day.
Missed connections, crew knock-on and cargo, beyond one cost per cancelled flight.
How it is checked: Before you rely on it, we replay your last twelve months on your own records: your technical cancellations and ours, month by month, on one chart. The test is written in the order before you send any data; if we do not reproduce your cancellations, you don’t pay. Every run carries a fingerprint, printed on its paper and its memo, so a figure read in committee leads back to its run.
Your data: It comes under an NDA, yours or ours, signed before you send anything. Aggregated, no personal data, no tail numbers unless the study needs them. Kept on our server in the European Union (Hetzner, Helsinki) and the founder’s encrypted computer, on no third-party service without your written consent. Used for your study only, never pooled, never used to train a model, and deleted within thirty days of the study’s end, the deletion confirmed in writing.

Every input of the sample run with its source, and its limits: the full memo (PDF), annexes B and C.