A freight forwarder books a shipment on a flight scheduled to land at 14:30. They tell the warehouse to expect it at 15:00, accounting for taxi and unload. The flight actually lands at 15:40 because of a ground stop at the origin, headwinds across the Atlantic, and a hold for traffic at arrival. The warehouse crew stood around for 40 minutes, the truck that was supposed to pick up the cargo left and won't be back for two hours, and the shipment misses its onward connection.
This happens constantly. It happens because the scheduled arrival time — the one in the airline's booking system, the one the forwarder passed downstream — was set days or weeks before departure and was never meaningfully updated.
Why the scheduled time is wrong so often
Airline schedules are built with block-time padding that's supposed to absorb normal variation. On average, they do. But logistics doesn't care about averages; it cares about the specific flight carrying the specific shipment today.
Three things make the scheduled time unreliable for any given flight:
Weather at origin or destination. A ground delay program at JFK adds 45 minutes to every departure. That information is public within minutes of being issued, but it doesn't flow into the freight tracking system automatically. The scheduled time sits there unchanged.
Winds aloft. Westbound transatlantic flights routinely run 30–60 minutes longer than eastbound flights on the same route, depending on the jet stream. The schedule uses an average block time that splits the difference. On a strong-jet-stream day, the westbound flight is going to be late by the schedule's definition, and everyone downstream is surprised.
ATC and traffic. Arrival sequencing at busy airports adds holding time that's impossible to predict at departure but becomes visible once the aircraft is in the terminal area. An aircraft entering a hold at FL100 is going to land 15–20 minutes later than one cleared for the approach, and ADS-B shows you which case you're in.
The fix for all three is the same: stop using the scheduled time once the flight is airborne and start using the actual position of the aircraft.
From position to ETA
A live ADS-B position gives you latitude, longitude, altitude, ground speed, and heading. That's enough to compute a rough ETA yourself: great-circle distance to the destination divided by current ground speed. It's better than the schedule, but it's also wrong — it doesn't account for the descent profile, arrival sequencing, or the fact that ground speed drops as the aircraft descends.
SkyLink's flight time prediction endpoint does this better by combining the live position with historical block-time data for the route pair. The output is a predicted arrival time and remaining flight minutes. It won't capture a holding pattern that hasn't started yet, but it will give you a prediction that improves steadily as the flight gets closer to landing.
import requests
BASE = "https://skylink-api.p.rapidapi.com"
H = {"X-RapidAPI-Key": KEY, "X-RapidAPI-Host": "skylink-api.p.rapidapi.com"}
# Get predicted arrival for a specific flight
r = requests.get(f"{BASE}/ml/flight-time", headers=H,
params={"flight_iata": "BA178"})
pred = r.json()
print(f"Predicted arrival: {pred.get('predicted_arrival')}")
print(f"Remaining minutes: {pred.get('remaining_minutes')}")
print(f"Model version: {pred.get('model_version')}")The model_version field matters for operational use. If the ML model is temporarily unavailable, the endpoint falls back to a simpler distance/speed calculation and tells you so through that field. A logistics system should treat a fallback prediction differently — it's still better than the schedule, but the confidence interval is wider.
Building this into a logistics pipeline
The typical freight-tracking architecture polls for updates and pushes them to downstream consumers — the warehouse management system, the trucking dispatcher, the customer-facing tracking page. Here's what the integration looks like for air legs:
Pre-departure: Pull the flight status to confirm the flight hasn't been cancelled or rescheduled. The flight status endpoint returns the current status, departure gate, and any delays already assigned. If the origin airport has a ground delay program, the FAA delays endpoint will tell you.
# Check FAA delays at origin
delays = requests.get(f"{BASE}/delays/faa", headers=H)
for d in delays.json().get("data", []):
if d.get("airport") == "JFK":
print(f"JFK delay: {d.get('reason')} — avg {d.get('average_delay')}")In-flight: Poll the aircraft's position every 60–120 seconds. Each poll updates your ETA. Push the updated ETA downstream whenever it shifts by more than 5 minutes from the last one you sent — this avoids flooding downstream systems with noise from minor fluctuations.
On approach: Once the aircraft is below 10,000 feet and within 30 miles of the destination, your ETA is probably accurate to within 5 minutes. This is when the warehouse should stage the unload crew and the truck dispatcher should confirm the pickup.
On the ground: is_on_ground flips to true. The aircraft still needs to taxi to the gate, which takes 2–20 minutes depending on the airport and the landing runway. You won't get a precise gate-in time from ADS-B, but you can add a fixed taxi buffer based on the airport.
# Poll for a specific cargo flight
r = requests.get(f"{BASE}/adsb/aircraft", headers=H,
params={"callsign": "CLX789"})
ac = r.json().get("aircraft", [])
if ac:
a = ac[0]
if a.get("is_on_ground"):
print(f"Landed — taxiing at {a.get('ground_speed')} kts")
else:
print(f"Airborne — alt {a.get('altitude')} ft, "
f"gs {a.get('ground_speed')} kts")The business case in plain numbers
The value isn't in the ETA itself — it's in what everyone downstream does with a better estimate.
A warehouse crew staged 30 minutes early is 30 minutes of labour wasted. Multiply that by several flights a day across a network of stations and you're looking at hundreds of hours per month. A truck that shows up on time instead of waiting or making a second trip saves fuel and a driver's time.
The bigger number is usually the missed connection. A shipment that misses its onward flight by 10 minutes because nobody updated the handling priority sits in a warehouse until the next available flight, which might be tomorrow. For time-sensitive freight — pharmaceuticals, fresh goods, just-in-time manufacturing parts — a day of delay has a cost that dwarfs the API subscription.
What you need to get started
The endpoints that matter for logistics ETAs are the ADS-B position feed, the ML flight time prediction, and the flight status endpoint. All three are available on the free trial with 1,000 requests/month — enough to prototype the polling loop for a handful of flights and validate the ETA accuracy against actual arrival times before committing to a production integration.
For production logistics platforms tracking hundreds of flights per day, the volume pricing is on the free trial page, or talk to us about Enterprise if you need an SLA behind it.