Where Flight Tracking Data Comes From: Feeds, Networks and APIs

Zoom in on any airliner icon crawling across a tracking map and you are looking at a small miracle of logistics, not one signal but a braid of them. The position may have arrived via a retired engineer's rooftop antenna in Ohio. The callsign came from the aircraft itself, but the route, the aircraft photo, the scheduled arrival time and the gate came from entirely different pipelines, some corporate, some governmental, some crowd-sourced. Flight tracking looks like a single feed and is actually a supply chain, and like all supply chains it has tiers, bottlenecks, redundancies and politics. Trace one icon back to its origins and you learn how the entire industry works.

Tier one: The radio layer

Everything starts with radio-frequency truth, the signals aircraft actually emit. The richest source is ADS-B, the self-reported GPS position broadcast on 1090 MHz that we dissect in our ADS-B guide; it supplies identity, position, altitude and velocity, refreshed about twice a second. Where aircraft transmit without position, receiver networks compute location by timing, the multilateration technique that requires four stations to hear the same message. In the United States, a second broadcast flavor, 978 MHz UAT, carries much general aviation traffic. And over oceans, satellite-hosted receivers, led by the Aireon payloads on Iridium NEXT that we profile in our satellite tracking article, catch the same 1090 MHz messages from 780 kilometers up.

Who owns the antennas matters enormously. The major commercial trackers each cultivate networks of tens of thousands of volunteer-hosted receivers, seeded by giving free hardware or premium subscriptions to hosts in coverage gaps. Independent community networks pool hobbyist feeds under open or semi-open terms. And the academic world has its own: the OpenSky Network, founded in 2013 by researchers, collects raw Mode S and ADS-B data from thousands of stations and preserves the unfiltered record for science, hundreds of billions of messages that have fed studies on everything from collision-avoidance behavior to pandemic traffic collapse. If hosting a station yourself sounds appealing, our walkthrough on building a Raspberry Pi ADS-B receiver shows how cheap the on-ramp is.

Tier two: Government and ANSP feeds

The second tier comes from the people who run the airspace. The FAA publishes a rich data service through its System Wide Information Management program, SWIM, which streams surveillance-derived traffic data, flight plans, airport events and weather products to approved consumers; details live on the FAA SWIM pages. Its traffic feed descends from the older ASDI service that powered the first generation of US flight tracking websites in the 1990s, long before hobbyist ADS-B existed. Government feeds have two special properties: they include radar-derived targets that no hobbyist can receive, and they honor official blocking requests, delivering data delayed by several minutes and omitting aircraft whose operators have opted out.

Other regions are stingier. European surveillance data is coordinated through Eurocontrol, whose Network Manager data reaches airlines and airports but is not broadly published in real time. This asymmetry explains a pattern sharp-eyed users notice: some aircraft visible over the US on certain services vanish over Europe, and vice versa, because the underlying entitlements differ. Where a tracker's coverage comes from matters as much as how much coverage it has.

Tier three: The context layer

Radio signals tell you a transponder is at a point in space wearing a callsign. Everything else on the flight's detail page is context, merged from separate sources. Schedules and route data come from airline filings and distribution systems coordinated through industry bodies like IATA; they are what let a tracker say this is flight BA117, London Heathrow to New York JFK, due at 13:40. Registration databases, national civil registries like the FAA's, plus commercial fleet data, translate the hex code into a specific airframe with an age, an owner and a seat map; the mechanics of that translation run through the 24-bit addressing scheme explained in our Mode S article. Airport status, gates and delays flow from airport operator systems. Photographs come from spotter communities. Weather overlays come from meteorological agencies such as NOAA.

Stitching these together is harder than it sounds because the keys do not naturally match. The radio layer speaks hex codes and callsigns; the schedule layer speaks flight numbers and city pairs; the registry layer speaks tail numbers. A single flight might broadcast callsign BAW117 while selling tickets as BA117 and four codeshare numbers, and the mapping between callsign and flight number is convention, not law, a tangle we unravel in our basics piece on callsigns versus flight numbers. Every tracker maintains a matching engine, and every tracker's engine occasionally guesses wrong, which is why you sometimes see an aircraft labeled with yesterday's route.

A tracking map is not a window onto the sky; it is an editorial product, assembled from a dozen feeds, each with its own owner, latency and blind spots.

The fusion engine: From messages to map

Now consider the merge. A busy tracker ingests tens of thousands of position messages per second, with the same aircraft often reported simultaneously by fifty receivers, an MLAT solver, a satellite feed and a government stream. The fusion pipeline deduplicates by hex code and timestamp, weighs sources by accuracy, ADS-B usually wins, then MLAT, then delayed radar feeds, rejects impossible jumps, smooths jitter, and interpolates between updates so icons glide instead of stutter. When all live sources go quiet, many services keep the icon moving on a dead-reckoned estimate along the filed route, marked as estimated, a courtesy that becomes obvious over mid-ocean gaps. The output is republished to map servers, mobile apps, alert engines and the live map you can open right now.

Then comes distribution. Nearly every major tracker sells the fused product back out through commercial APIs, consumed by airlines watching their own fleets, airports feeding arrival boards, logistics firms timing cargo connections, financial analysts counting flights as economic signals, and journalists reconstructing incidents. Meanwhile OpenSky offers researchers free access to historical raw data, a resource that has appeared in over a hundred peer-reviewed papers. Aviation data, once a byproduct, is now a proper industry with tiered pricing, licensing disputes and genuine public-interest stakes.

How to judge what you are seeing

All of this yields a practical literacy for map readers. Check the position source field on a flight's detail panel: ADS-B means GPS-grade truth from the aircraft, MLAT means computed timing with modest wobble, estimated means educated fiction along a filed route. Notice the update cadence: sub-second smoothness suggests dense terrestrial reception, while stately several-second hops over the Pacific suggest satellite relay. Be suspicious of context fields, routes and gates especially, which are the most frequently wrong because they travel the longest, most human pipeline. And when two tracking services disagree about the same flight, you are not seeing an error so much as two different supply chains delivering two different editorial products. For the raw demographic behind it all, thousands of aircraft aloft at any moment, our statistics page aggregates the fused picture and connects the callsigns you decode to the carriers flying them.

The next time an icon glides across your screen, give a thought to the braid behind it: a rooftop antenna, a timing server, a satellite crosslink, a schedule file and a matching engine, all converging on one small moving symbol that happens, several seconds and several handoffs later, to be true.