Frequently asked questions

Flightradar24 subscribers can download a flight's track as a CSV file from the flight's playback page. flightanalyser reads that seven-column format: Timestamp, UTC, Callsign, Position, Altitude, Speed, Direction. Save the file, then drop it onto the upload area on the Analyse page, or choose it with the file picker. The file is analysed in your browser and is yours to use under your Flightradar24 subscription.

flightanalyser never asks for a login to another service and does not fetch the file for you. Without an export, a flight lookup or the sample flight will load a flight instead.

Exactly seven columns, in this order, under one header row:

Timestamp,UTC,Callsign,Position,Altitude,Speed,Direction
1768467600,2026-01-15T09:00:00.000Z,DEMO123,"52.310000,4.750000",0,0,205

• Timestamp is Unix epoch seconds and is the authoritative time: rows are sorted on it, whatever order the file uses. UTC is for display only.
• Position is a single quoted field holding latitude and longitude, separated by a comma.
• Altitude is barometric altitude in feet, Speed is groundspeed in knots, and Direction is track in degrees true.

Any other header is rejected with a message naming the expected columns. Rows outside these bounds are dropped and counted rather than failing the file: latitude ±90°, longitude ±180°, altitude −2,000 to 70,000 ft, speed 0 to 1,500 kt, direction 0 to 360°. Duplicate timestamps are removed. Files over 10 MB or 100,000 rows are refused before any row is parsed.

Not unless you save it. An uploaded CSV is read and analysed entirely in your browser. It is not sent to the flightanalyser server and not stored. The analysis exists only in the memory of the open tab: closing or reloading the tab discards it.

On the Light and Pro plans you can save an upload to your account, to reopen it later on any device. Only then is the file sent and stored, until you delete it.

The PDF report is generated in the browser too. Something about an uploaded flight leaves your browser only if you start the route benchmark or the operating pattern: then the two airport codes, or the flight number, are sent to the server to retrieve recorded flight data. The track itself is not.

A synthetic London Heathrow–New York JFK sector, produced by simulation and written in the seven-column CSV format so that one load demonstrates every analysis. The simulated track includes a taxi with a stop at the holding point, a level-off in the climb, two step climbs, a coverage gap in cruise, a level-off in the descent, holding circuits before the approach, and a sunset in flight followed by a night arrival.

It is not a real flight.

The file is split into sessions wherever consecutive points are more than 20 minutes apart. You choose which session to analyse; the default is the longest session in which the aircraft flew.

One exception is deliberate. Long-haul exports routinely lose coverage for 30–50 minutes over oceans and remote land, and splitting there would cut one flight in two. A gap of more than 20 minutes is therefore not split when all of these hold:

• the aircraft is clearly airborne on both sides of the gap (above 1,500 ft and 100 kt);
• the callsign is unchanged;
• the distance covered implies continuous flight at roughly the speeds either side, and the gap is no longer than four hours.

A gap that contains a landing and a turnaround fails the speed test, so a file holding genuinely separate flights still splits. A bridged gap is then treated like any other gap: interpolated, flagged, and reflected in the confidence score.

Only to look flights up. Uploading a CSV track export and opening the sample flight need no account. A free account includes 1 flight lookup a month; Light and Pro include more, and the extra lookups such as departure traffic and the route benchmark.

Allowances renew every month on the date your paid plan started, or the date you joined on the free plan: a plan started on 20 October renews on 20 November, then 20 December. A flight counts once until the allowance renews: opening it again uses no extra lookup, and a lookup that fails is not counted. Searching for a flight is free.

Flight lookup is occasionally unavailable; CSV upload always works.

A flight that is still in the air can be looked up once it has landed; until then its recorded track is incomplete.

No. flightanalyser analyses one flight at a time, in depth; loading another flight replaces the current one. For context, the route benchmark places the loaded flight against up to 60 other sectors on the same route from the last 30 days.

Observed figures are read from the recorded track or computed from it by simple arithmetic: times, durations, distances, altitudes, groundspeeds and sampling statistics.

Analysed figures are worked out by flightanalyser: the efficiency score and findings, holding patterns and step climbs, cabin-side sun and red-eye status, benchmark verdicts, and airports resolved by proximity to the track ends. They depend on thresholds and reference values that the track itself does not contain.

Reported figures come with a flight lookup — registration, Mode-S address, operator, aircraft type and runways — and are shown as received. Each kind is labelled wherever it appears.

From altitude, groundspeed and movement, in a fixed order: the airborne span first, because it is the most robust signal, then the ground events either side of it, then cruise within it.

• Airborne means more than 50 ft above the ground elevation measured from the stationary points at each end of the track, and faster than 80 kt.
• Taxi-out starts with sustained ground movement above 5 kt; the takeoff roll starts with the acceleration through 40 kt.
• Cruise is settled level flight above 18,000 ft, found on a median-filtered altitude so brief level-offs and noise are ignored.
• Approach begins below 5,000 ft above the arrival airport.

Nine events anchor every duration: taxi-out start, takeoff roll start, takeoff, top of climb, descent start, approach start, landing, taxi-in start and final stop. Taxi-in runs from below 35 kt after touchdown until groundspeed stays below 1 kt. If the recording ends while the aircraft is still taxiing, taxi-in runs to the last recorded position and is marked approximate; if the track never slows below 35 kt, it reads N/A. Where a track starts or ends in the air, the events it does not cover are left blank rather than guessed.

Because a single corrupted ADS-B sample would otherwise define the headline figure. Recorded tracks occasionally carry one physically impossible groundspeed; the 99th percentile sets aside the top 1% of observed airborne samples, so one bad value cannot set it, while a genuinely sustained high speed still shows.

Only observed airborne groundspeeds count: interpolated points and samples whose speed was replaced by a position-derived estimate are excluded. The figure is groundspeed, not airspeed, so a strong tailwind raises it — a jet-stream crossing can genuinely exceed 600 kt.

A gap is an interval longer than three times the track's normal sampling cadence, and at least two minutes. Across each gap, synthetic points are placed on the great circle between the observed points either side, 15–45 seconds apart and never more than 180 per gap, so the charts and map stay continuous without a long gap looking like real coverage.

Synthetic points are flagged. They lower the confidence score, are excluded from speed statistics and holding detection, and cannot by themselves place a takeoff or landing. The longest gaps each produce an analysis note naming the phase and what it makes approximate — for example, “4-minute gap during approach; approach timings are approximate.”

Above 1,500 ft, each sample's reported groundspeed and track is compared with the aircraft's actual movement between neighbouring positions. Where they disagree, the reported value is replaced by a position-derived estimate for phase detection, holding detection and cabin-side sun, and the sample is excluded from speed statistics. An analysis note reports how many samples were affected.

This exists because of GNSS interference. Flights crossing the Gulf in 2026 have been recorded with groundspeed swinging between 0 and 950 kt at cruise level and track pointing in random directions, while the positions moved smoothly at around 480 kt. Left alone, those values would set the maximum speed, break the airborne test and invent holding circuits.

Isolated position spikes — a single point implying more than 1,000 kt that then returns to the track — are removed outright.

It rates the data, not the flight: how far the analysis can be trusted given the track it was built from. It runs from 0 to 100 in three bands — high (80 and above), medium (55–79) and low (below 55).

Points are deducted for the longest gap and the number of gaps, the share of interpolated points, irregular sampling, sparse airborne coverage, speed or track inconsistent with position, a track that starts or ends in the air or while taxiing, an estimated takeoff roll, missing altitude, and rows that failed validation. The analysis notes then say what each issue affects, such as “taxi-in and block time are approximate”.

All dates and times are UTC unless labelled local. Local clocks for departure and arrival use each airport's time zone, derived from its coordinates, with the UTC offset in force at that moment; the time-zone shift is the difference between the two offsets. Near a border, a derived zone can occasionally be the neighbouring one.

A sector counts as a red-eye when it departs between 21:00 and 02:00 local, arrives between 04:00 and 10:00 local, spends at least three hours airborne, and has at least 40% of its airborne time in twilight or darkness.

Daylight uses the standard solar elevations: sunrise and sunset at −0.833°, civil twilight down to −6°. Sun on the left or right of the cabin comes from the bearing to the sun relative to the aircraft's track; sun within 10° of the nose or tail counts for neither side.

It is an analysed figure. The score starts at 100 and each of five findings deducts a capped penalty, so the grade always reconciles with the findings that produced it:

• Route distance — up to 25 points, for distance beyond a routing allowance of 3% of the great circle plus 30 nm.
• Holding — up to 20 points, by minutes in the hold.
• Taxi-out — up to 15 points, for time beyond a general 15-minute reference.
• Descent — up to 12 points, for level segments of a minute or more above 5,000 ft over the arrival airport.
• Step climbs — up to 6 points when none is flown, assessed only on flights airborne for more than three hours.

Grades: A at 90 or above, B at 80, C at 70, D at 60, otherwise E. Each finding is rated good, minor, notable or significant; one the track cannot support, such as taxi-out on a track that starts airborne, is marked not assessed and deducts nothing. The references are general: the taxi reference is not the airport's own unimpeded time. Cruise level is not scored, because the optimum level depends on weight, temperature and the operator's cost index, none of which are in the track.

Because the great circle is rarely a route an aircraft can fly. Some extension is structural: airways, departure and arrival procedures, restricted or closed airspace, and spacing by air traffic control. A wind-optimal route can also be longer than the great circle yet faster — on a jet-stream day, flying further to ride a tailwind is the efficient choice.

That is why the score counts only the distance beyond the routing allowance, and why the finding is worded as a prompt to look, not proof of inefficiency. The track shows what was flown, not the reason for it.

It is optional and runs only when you start it, for any loaded flight whose departure and arrival airports are resolved — an upload included.

It samples up to 60 completed sectors on the same route from the last 30 days and reports p10, median and p90 for block time, airborne time, route extension, taxi-out and taxi-in, with the loaded flight's percentile and a plain-language verdict. Live and incomplete rows are excluded and counted; diversions are excluded from the distributions but reported, with the diversion rate. Placements are withheld when fewer than five sectors are usable.

The benchmark's block and taxi figures use first and last ADS-B contact as a proxy, and the loaded flight is measured the same way for the comparison, so they will not match the phase-based taxi times elsewhere in the analysis.

A Monday-to-Sunday calendar of the days the flight number was observed in flight records over the last 28 complete UTC days, with a written summary such as which weekdays it regularly operated. It is observed, not a published schedule: a blank day may be a cancellation, a coverage gap or a renumbered flight. It is optional and runs only when you start it.

Because the analysis sees only the recorded surveillance track, and measures it in its own way. The main mechanisms:

• Block time here runs from the start of sustained taxi to the final stop. An airline's block time runs from off-blocks, often the start of pushback, to on-blocks, so the two can differ by a few minutes.
• Coverage gaps, especially on the ground and over oceans, make event times approximate; the analysis notes say which.
• Altitude is the barometric altitude broadcast by the transponder, and speed is groundspeed, not the airspeed in the aircraft's own records.
• Great-circle distance runs between airport reference points, not runway ends, and an airport resolved by proximity to the track can occasionally be the wrong one.
• Phase boundaries come from fixed thresholds, and the efficiency references are general figures, not the operator's flight plan, weights or winds.
• Benchmark block and taxi figures use first and last ADS-B contact as a proxy.

No. flightanalyser is for general information only. Its figures are estimates from recorded ADS-B data, which can be incomplete or wrong, and its analysed values rest on general reference figures. It is not suitable for flight planning, dispatch, or any operational or safety-critical use, and nothing in it should be relied on for those purposes.

The whole trip from the moment the aircraft starts moving at the departure gate to the moment it stops at the arrival gate, including taxiing on the ground. Here it runs from sustained taxi movement to the final stop.

From takeoff (the wheels leaving the runway) to touchdown. Block time minus the time spent taxiing and on the runway.

The shortest path between two places on the Earth's surface. Flights rarely follow it exactly: airways, winds, airspace and departure and arrival procedures add distance.

TOC, top of climb, is where the climb ends and the cruise begins. TOD, top of descent, is where the descent from cruise begins.

A climb from one cruise altitude to a higher one during the flight, usually as fuel burns off and the lighter aircraft can fly higher more efficiently.

A settled level is a cruise altitude held long enough to count as a planned cruise level. A transient level is a shorter level-off, such as a temporary restriction from air traffic control.

Flying a racetrack-shaped pattern while waiting for clearance to continue, usually near the arrival airport when it is busy.

Speed over the ground, not through the air: a headwind lowers it and a tailwind raises it, so it can differ a lot from the aircraft's airspeed.

Maximum take-off weight: the heaviest the aircraft type is certified to be at takeoff, including passengers, cargo and fuel.

The aircraft's unique 24-bit transponder code, shown in hexadecimal. It belongs to the airframe and normally stays the same for its whole life.

Automatic Dependent Surveillance–Broadcast: the aircraft broadcasts its satellite position, altitude and speed every second or so, and receivers on the ground record it. Flight tracks come from these broadcasts.

Jamming or spoofing of satellite navigation signals (GPS and similar). It can make an aircraft report wrong positions or speeds for a while.

Coordinated Universal Time, the single time standard aviation uses worldwide. It does not change with time zones or summer time; London is on UTC in winter.

Aviation's units. A nautical mile is 1.852 km; a knot (kt) is one nautical mile per hour, 1.852 km/h; feet per minute (fpm) measures climb and descent. You can switch to kilometres, miles or metres in Settings.

Observed figures are seen directly in the aircraft's recorded track. Analysed figures are worked out by flightanalyser from it: estimates, not operator records. Reported figures come with a flight lookup, as stated by the flight data source.