Methodology

Airport eligibility

We publish a /data/airports/{IATA}/ page for a fixed selection of major airports.

Smaller airports are intentionally excluded - we'd rather skip them than publish thin, low-confidence data.

Hourly busyness

For each airport we count scheduled departures (and separately, arrivals) in each local-time hour bucket over a rolling 60-day window. Each aircraft movement counts once: a departure sold under several codeshare flight numbers is one departure. The displayed number is the average flights per day for that hour - dividing by the number of distinct days we observed the airport, so a half-month dataset doesn't get artificially halved and an hour with a single flight in the whole window reads as a fraction, not as one an hour.

• Peak hour: the single hour bucket with the highest average.
• Quietest active hour: the lowest non-zero bucket (so we don't always pick 03:00 at airports with no overnight ops).
• Weekday vs weekend: weekday = Mon-Fri local, weekend = Sat-Sun local at the airport's own timezone.

Top operators and destinations

Both are simple counts: the top airlines (by published flight number) and top other-airports (by paired destination/origin) observed during the same 60-day window. We:

• Count marketed flight numbers, so a codeshare counts once per marketing carrier. (A flight operated by Lufthansa but sold as a United codeshare counts on both lists - this is intentional and matches how passengers experience the schedule.)
• Trim airline names for display ("United Airlines, Inc." -> "United Airlines") but keep the IATA code in parentheses so identity is unambiguous.

Refresh cadence

• Schedule observations: continuously, as flights operate.
• Which airports get a page: rechecked every 6 hours.
• Flight pages: computed on request from the last 90 days of arrivals we hold.

Flight number pages

• A flight number gets a page once we have seen it operate often enough in the last 90 days, across every leg it flies. Punctuality is printed only once enough of those carry a confirmed arrival time.
• On time means the aircraft arrived no more than 15 minutes after schedule. The figure is rounded to the nearest 5 points, lateness to the nearest 5 minutes, and the sample to the nearest 10 flights. Cancellations are given as a "1 in N" bucket, not a rate.
• Where a number flies more than one leg (Sydney to Singapore to London, for example), punctuality is for the busiest leg and the page says which.
• Rounding is deliberate. These pages describe a pattern; the precise, live view of a flight is reserved for signed-in travellers tracking their own booking.

Known limits

• Coverage is uneven - some airports have months of data, some have weeks. Figures on the airport pages are rounded.
• We deliberately do not publish per-flight live status on /data/ - those views are reserved for signed-in users tracking their own trips.
• Historical averages are not a forecast. Holidays, weather, and ATC events can move a real day's traffic far away from the published average.
• "Quietest" rankings exclude very small airports, to avoid ranking low-volume noise.

Reuse

These pages are free to read, link to, and quote with attribution. A backlink to the originating page is appreciated.