How we measure

Every number on this site comes from flights Flight Owl tracked. This page says exactly what that means, and where it falls short.

What the data is, and is not

In plain terms: we can only measure the flights our users chose to follow. An airport where our users happen to fly at busy times will look worse than one where they fly at quiet times, even if the two airports are the same. Read these numbers as a well-evidenced impression, not as an official statistic.

The destination lists have the same shape of limit. They record routes we have observed, from user lookups, flight history and airport departure boards. A route missing from a list means we have not seen it, not that it does not exist.

The Delay Index

0 = calm, 100 = severely disrupted. Combines typical departure delay, how often a departure is 60+ minutes late, and the airport's average board disruption rate.

It combines three measurements, weighted:

• 40% typical departure delay, measured against a 60-minute ceiling.
• 30% how often a departure is 60 minutes or more late, measured against a 25% ceiling.
• 30% the share of the departure board delayed or cancelled, measured against a 50% ceiling.

Every ceiling except the 25% tail-risk one is a threshold Flight Owl already uses inside the app to call a delay severe, so the scale means the same thing here as it does there. The tail risk ceiling is our own editorial choice.

When we show nothing

A measurement we cannot make well, we do not publish. Each of the three parts has a minimum amount of evidence, and the index itself needs all three: if any part is missing we show no index at all, rather than quietly scoring an airport on the parts that happened to survive.

That is why most airports here have destinations but no index. We publish an airport page when we have seen at least 20destinations served from it, which is our editorial floor for a page being worth reading. 671 airports currently clear it. Once a page exists we keep it while the airport still serves 15, so a single quiet week does not take a page down and put it back up again.

Route and flight pages

A route page covers one origin and destination pair; a flight page covers one flight number on one of those pairs. Both need at least 10 departures we tracked and resolved, which is enough to know the leg genuinely runs and to describe it: who flies it, how often, how far, and what aircraft we have seen on it. 1,499 routes and 1,396 flights clear that floor.

Publishing a typical delay takes far more, and it is the same bar an airport has to clear: 30 departures spread across 30 separate days. A page can therefore exist with plenty to say about the schedule and nothing we are willing to claim about lateness, in which case the delay section is simply absent rather than empty.

A flight number is not a unique name. The same number is often used for a return leg, for several legs of one aircraft's day, or for an unrelated route on different days, so each page covers a number on a route and says so in its title. Aircraft types are only ever ones we recorded on that exact flight; we never borrow the type from another flight on the same route.

Why there are no dates on these pages

Every figure is a median or a share measured over months, so it does not change day to day. Nothing here is a forecast, and nothing describes a particular flight. For live flight status, delay predictions and gate alerts, that is what the app is for.