Skip to content
Public-source data

US Flight Delays Dataset

Analyze real on-time performance for every 2025 flight between the eight busiest US airports. Each flight records its schedule, actual times, delay, cancellation or diversion, and the reported causes of delay.

Good for: delay and cancellation rates, route rankings with window functions, and handling missing values.

Download this dataset

Pick the format that fits your project. Downloads are free with an account, and you can create one in the same step.

While it downloads: answer 5 questions in your browser

Run SQL on these tables right here, with no setup. Each answer is checked instantly, and every question comes with hints.

Preview the data

Explore a sample before downloading. Select a table to see its columns, row count, relationships, and what each row represents.

Airlines

One row represents: One row per reporting carrier.

Rows in this table
11
Connects to
Flights

Sample rows · download for the full table

Sample rows from airlines. This preview is not the full download.
airline_codeairline_name
AAAmerican Airlines Inc.
ASAlaska Airlines Inc.
DLDelta Air Lines Inc.
F9Frontier Airlines Inc.
MQEnvoy Air

Airlines data dictionary

Check identifiers, dates, numeric fields, and missing values before writing a query or building a chart.

ColumnWhat it meansData typeMissing valuesKey or requirement
airline_codeBTS unique carrier code of the airline that operated and reported the flight.text0Required
airline_nameCarrier name from the BTS carrier lookup table.text0Required
View the full data dictionary

airlines

One row represents: One row per reporting carrier.

ColumnWhat it meansData typeMissing valuesKey or requirement
airline_codeBTS unique carrier code of the airline that operated and reported the flight.text0Required
airline_nameCarrier name from the BTS carrier lookup table.text0Required

airports

One row represents: One row per airport.

ColumnWhat it meansData typeMissing valuesKey or requirement
airport_codeThree-letter airport code.text0Required
airport_nameAirport name. Added by Analytics Engineering; the BTS file carries only the code and city.text0Required
city_nameCity from the BTS record.text0Required
state_codeTwo-letter state code.text0Required
state_nameState name.text0Required
departures_all_routesFlights that departed this airport for any destination in the 2025 source files, including airports outside this dataset. This count is why the airport was selected.integer0Required

cancellation_codes

One row represents: One row per cancellation code.

ColumnWhat it meansData typeMissing valuesKey or requirement
cancellation_codeA, B, C, or D.text0Required
cancellation_reasonCarrier, Weather, National Air System, or Security.text0Required

flights

One row represents: One row per scheduled flight.

ColumnWhat it meansData typeMissing valuesKey or requirement
flight_idKey assigned by Analytics Engineering in order of flight date and scheduled departure. Not a BTS field.text0Required
flight_dateScheduled departure date at the origin airport.date0Required
airline_codeReporting carrier.text0Required
flight_numberFlight number.integer0Required
tail_numberAircraft tail number. Missing for some cancelled flights.text800Nullable
origin_airportOrigin airport code.text0Required
dest_airportScheduled destination airport code.text0Required
scheduled_departureScheduled departure, local time at the origin, as hh:mm.text0Required
actual_departureActual gate departure, local time, as hh:mm. Midnight can appear as 24:00. Missing for most cancelled flights.text3,827Nullable
departure_delay_minutesActual minus scheduled departure in minutes. Negative when the flight left early.integer3,830Nullable
taxi_out_minutesMinutes from gate departure to wheels off.integer3,961Nullable
taxi_in_minutesMinutes from wheels on to gate arrival.integer3,991Nullable
scheduled_arrivalScheduled arrival, local time at the destination, as hh:mm.text0Required
actual_arrivalActual gate arrival, local time, as hh:mm. Midnight can appear as 24:00. Missing for cancelled flights and for a few diverted ones.text3,991Nullable
arrival_delay_minutesActual minus scheduled arrival in minutes. Negative when the flight arrived early. Missing for cancelled and diverted flights.integer4,754Nullable
cancelled1 when the flight was cancelled, otherwise 0.integer0Required
cancellation_codeReason code when cancelled, otherwise missing.text310,654Nullable
diverted1 when the flight landed somewhere other than its scheduled destination, otherwise 0.integer0Required
scheduled_elapsed_minutesScheduled gate-to-gate time in minutes.integer0Required
actual_elapsed_minutesActual gate-to-gate time in minutes. Equals taxi out plus air time plus taxi in.integer4,754Nullable
air_time_minutesMinutes in the air.integer4,754Nullable
distance_milesDistance between the two airports in miles.integer0Required
carrier_delay_minutesMinutes of arrival delay attributed to the carrier. The five cause columns are filled only when the flight arrived 15 or more minutes late, and they add up to the arrival delay.integer242,153Nullable
weather_delay_minutesMinutes of arrival delay attributed to extreme weather.integer242,153Nullable
nas_delay_minutesMinutes of arrival delay attributed to the National Air System, such as air traffic control, heavy traffic, and weather that is not extreme.integer242,153Nullable
security_delay_minutesMinutes of arrival delay attributed to security.integer242,153Nullable
late_aircraft_delay_minutesMinutes of arrival delay because the aircraft arrived late from an earlier flight.integer242,153Nullable

About the data

This dataset is real. It contains the 314,627 scheduled flights that flew between two of the eight US airports with the most departures in 2025, selected from the on-time performance data that airlines report to the US Bureau of Transportation Statistics. Nothing is simulated or filled in.

Real data comes with rules you have to learn. A cancelled flight has no arrival time. A diverted flight has no arrival delay. The five delay-cause columns are filled only when a flight arrived 15 or more minutes late. Getting a correct average means deciding what to do with each of those cases.

What one row represents
Varies by source table
Data covers
Start: 2025-01-01 · End: 2025-12-31
Dataset version
1.0.0
Release published
2026-11-22

How the tables connect

Generated from the same relationship metadata as the table documentation.

Flightsairline_codeAirlinesairline_code

many-to-one

Flightsorigin_airportAirportsairport_code

many-to-one

Flightsdest_airportAirportsairport_code

many-to-one

Flightscancellation_codeCancellation codescancellation_code

many-to-one

Start your first analysis

Open a download, inspect the table, and answer one question.

Start with Excel

Use the prepared one-row-per-flight workbook to compare months, airlines, and routes without joins.

Download Excel
File used
flight-delays-v1.0.0-analysis.xlsx
Dataset version
1.0.0
File size
24.1 MB
  1. 01Open "flight-delays-v1.0.0-analysis.xlsx" and select the "US Flight Delays Dataset" sheet. It has 314,627 rows, so give it a moment to load.
  2. 02Insert a PivotTable with month in Rows, status in Columns, and flight_id (count) in Values.
  3. 03Add arrived_15_or_more_minutes_late (average) to Values to see the late-arrival rate by month.

What you can build

Finished output from this release

Cancellations and late arrivals by month

Arrivals measured counts flights that have an arrival delay, which excludes cancelled and diverted flights. The late share is the percentage of those flights that arrived 15 or more minutes after schedule.

MonthScheduled flightsCancelledArrivals measuredArrived 15+ min late (%)
2025-01255136592481818.4
2025-02236981052357018.1
2025-03280912312781019.8
2025-04271392862678423.8
2025-05279071992764725.2
2025-06268894132637131.2
View the checked SQL
SELECT substr(flight_date, 1, 7) AS Month,
       COUNT(*) AS Scheduled flights,
       SUM(cancelled) AS Cancelled,
       COUNT(arrival_delay_minutes) AS Arrivals measured,
       ROUND(100.0 * SUM(arrival_delay_minutes >= 15) / COUNT(arrival_delay_minutes), 1) AS Arrived 15+ min late (%)
FROM flights
GROUP BY month
ORDER BY month;

On-time performance report

Report the share of flights arriving 15 or more minutes late and the cancellation rate by month and airline. State how many flights each rate is based on.

Route ranking

Rank the 56 routes by late-arrival rate within each origin airport using window functions. Compare both directions of the same route.

Delay cause breakdown

Reshape the five cause columns into rows and show which causes drive delay minutes in summer and in winter.

Aircraft rotation analysis

Follow each tail number through its day with LAG and measure how often a late arrival is followed by a late departure on the same aircraft.

Questions about this dataset

Is this real flight data?

Yes. Every row is a flight reported to the US Bureau of Transportation Statistics in its Reporting Carrier On-Time Performance data for 2025. The values are as reported. Only the column names and the way numbers and times are written were changed.

Why only eight airports?

The full 2025 source has just over 7 million flights, too many for a spreadsheet. This dataset keeps flights whose origin and destination are both among the eight airports with the most departures: Atlanta, Charlotte, Chicago O'Hare, Dallas/Fort Worth, Denver, Las Vegas, Los Angeles, and Phoenix. That gives 56 routes with at least 2,300 flights each.

Can I use it to say which airline is most reliable?

Not in general. These are 314,627 of about 7 million flights, on routes between large hubs, and each flight is listed under the carrier that operated it. Use it to practice the analysis, and use the full BTS data for conclusions about an airline.

Why are the delay cause columns empty for most flights?

Airlines report causes only for flights that arrive 15 or more minutes late. For those flights the five cause columns add up to the arrival delay. For every other flight they are missing, which is different from zero.

Can I use this data in a portfolio or commercial project?

Yes. It is a work of the US Government and is not protected by copyright in the United States. The Department of Transportation asks that you cite the Bureau of Transportation Statistics as the source.

Source, license, and versions

Before you interpret the results

  • Only flights whose origin and destination are both among the eight selected airports are included: 314,627 of the 7,001,619 flights in the 2025 source files. Rates here do not describe those airports' other routes or US aviation as a whole.
  • Covers only carriers that report on-time data to BTS, and each flight is listed under the carrier that operated it, which can be a regional partner of the airline that sold the ticket.
  • Arrival delay is missing for cancelled and diverted flights, and delay causes are recorded only for flights that arrived 15 or more minutes late.
  • Times are local to each airport and stored as hh:mm text, so a flight that crosses time zones or midnight needs care before subtracting times.
  • Alaska Airlines has 65 flights in this selection, too few to compare with other carriers.

Source, license, and version

Cite the version number when you share your work so anyone can rerun it on the same files.

Maintainer
U.S. Bureau of Transportation Statistics; selection and prepared formats by Analytics Engineering
License
Public domain (U.S. Government work)
Version
1.0.0
Published
2026-11-22T17:02:59.765Z
View the original source →

Cite this dataset

U.S. Department of Transportation, Bureau of Transportation Statistics. Reporting Carrier On-Time Performance (1987-present), 2025 flights. Prepared as US Flight Delays Dataset (Version 1.0.0) by Analytics Engineering (2026). https://www.analyticsengineering.com/datasets/flight-delays.

Found an issue?

Send the dataset, version, and file you used, plus what you expected to see. A small example gets it fixed faster.

Report a dataset issue →

Keep learning

Work with real operational data: missing values that mean something, three flight outcomes, and delay causes recorded only above a threshold.

Practice exercises run in your browser on a smaller sample of this dataset.