Methods and sources

Where every figure comes from, how each estimate is computed, and how far each estimate sits from what airlines report.

Sourcing rule

Open

Published by a public source and reusable: tax tables, airport charges documents, airline results, fee pages, seat maps, official statistics.

Derived

Built by AV8 from its own observations, accumulating over time: flights, operations, capacity, route records, change history.

Not used

Licensed data feeds, booking engines, internal APIs, logins and any information from an employer or a private source.

Every source document is archived before it is read, so a figure on a page can always be traced to the copy it came from. Where a source states no licence or reuse terms, the source is recorded and used; anything with a genuine legal flag is held back.

Basis labels

Every figure on a page carries one basis. In the API, meta.basis gives the basis of each field group.

Observed

Recorded by AV8 from public ADS-B data. Marked with a circle.

Published

Stated by a government, airport or airline. Marked with a square.

Estimated

Computed by a named AV8 method; the method name is given with the figure. Marked with a diamond.

Methods

Estimated ASK ask-gc-v1 Estimated

ASK = Σ over observed flights ( seats(airline, type) × great-circle km(departure, arrival) )

Inputs: Observed flights with both ends matched to an airport; the seat figure for the airline and aircraft type from the best available seat method.
Limits: Flown distance is longer than great-circle distance. Flights operated for another airline under this callsign are included. No seat figure: no estimate. Freighter flights are excluded.
Error: Ratio of the estimate to reported ASK, recorded every reported period (calibration-v1).

Flights and end matching legs-v1 Observed

A flight runs from take-off to landing in one aircraft's trace. Each end is matched to an airport by the first method that applies:
ground        last (first) ground position within the airport's radius
runway        low and aligned with a runway within 8 km
approach      descending (climbing) towards one large or medium airport within 25 km, below the approach (climb) line
extrapolated  projected along track and vertical rate to one large or medium airport within 30 km
quality A = both ends matched; B = one end; no end matched = not stored

Order: ground, then runway, then approach, then extrapolated. approach and extrapolated apply only where ground and runway find nothing, never within 5 minutes of 00:00 UTC, and never on MLAT positions. An end that fits two airports stays unmatched.
Measured: On 10 Oct 2026 (23,420 flights), against each callsign's published route where the other end is on that route: ground 0.49% not on the route, runway 0.77%, approach 1.83%, extrapolated 0.82%. Against the same aircraft's next or previous flight within 2 hours: approach 0 of 554 differ, extrapolated 1 of 91.
Record: The method of each end is kept as snap_dep and snap_arr in the open download, so the stricter set (ground and runway only) can be selected at any time.

Utilisation and stage length ops-v1 Observed

utilisation = Σ airborne hours ÷ aircraft observed ÷ 7
stage length = mean great-circle km per flight
turn time = median ground time between landing and the next take-off at the same airport, within 6 hours

Inputs: Observed flights with both ends matched, in complete weeks. Airborne time runs from take-off to landing.
Limits: Airborne hours are shorter than block hours, which include taxi time. Shown as airborne hours throughout.

Calibration calibration-v1 Estimated

ratio = ( AV8 estimate over observed weeks × days in period ÷ days observed ) ÷ reported figure

Scope: ASK and sectors. RASK, CASK and route margin are added in a later phase and calibrated the same way.
Record: Every period for every airline is kept and published, including periods where the ratio is poor.

No calibration ratios recorded yet.

Seat methods

Methods in order of preference per airline and aircraft type. Every seat figure states its method, source and date. ASK is Estimated whatever the seat method.

Method	Description	Basis	Exact
fleet_page	Airline fleet page	Published	Yes
seat_map	Airline seat map	Published	Yes
press_release	Airline press release	Published	Yes
reported_per_flight	Airline-reported seats divided by observed flights	Estimated	No
known_configuration	Known configuration from a cited public source	Published if the cited source is the airline, else Estimated	Yes or no
manufacturer_typical	Manufacturer typical seating	Estimated	No

Route records

Route records are counted from flights with both ends matched to an airport, in complete weeks only. A route change is not published when either airport's coverage that week is below 75% of its 28-day baseline. Wording is fixed:

First observed: {prefix} flights {DEP}–{ARR} in the week of {date} ({n} flights).
No {prefix} flights observed {DEP}–{ARR} since {date} (airport coverage {x}%).
{prefix} flights {DEP}–{ARR}: {a} in the week of {date}, {b} four weeks earlier.

For the first six months, every record of no flights and every lower weekly count is reviewed before it is published.