kvlak limn
DOT / BTS On-Time Performance July 4, 2026

On Time,
Honestly

The whole American departures board, read honestly. Every scheduled airline in the country files its arrival and departure times with the federal government, and the record is unusually blunt: a flight is on time if it lands less than fifteen minutes late, and not if it doesn't. This is a reading of that record - which airlines actually keep the promise, which routes are late by design, how the schedule itself gets padded to flatter the numbers, and the handful of days the whole board went dark. About 1.3% of flights never leave at all.

The Shape of a Delay

I. How late is a late flight? BTS data

Averages hide the story. The typical flight actually lands a few minutes early - the fat middle of this curve sits just left of zero. What matters is the right tail: the fifth of flights that cross the fifteen-minute line and become, officially, late. Most of those are only a little late. A stubborn few are late by hours.

0% 10% 20% the 15-minute line right of here = "delayed" 1.8% 30+ min early 1.8% of flights 18.6% 15-30 early 18.6% of flights 25.9% 5-15 early 25.9% of flights 18.5% within 5 min 18.5% of flights 10.2% 5-15 late 10.2% of flights 8.2% 15-30 late 8.2% of flights 7.5% 30-60 late 7.5% of flights 5.4% 1-2 hr late 5.4% of flights 3.7% 2 hr+ late 3.7% of flights 30+ min early15-30 early5-15 earlywithin 5 min5-15 late15-30 late30-60 late1-2 hr late2 hr+ late arrival vs. schedule - early on the left, late on the right
On time (< 15 min late) · Delayed · An hour or more late · 75% on time · 25% delayed · 9.2% over an hour
Every bin, in numbers
Arrival vs. scheduleShare of flightsStatus
30+ min early 1.8% on time
15-30 early 18.6% on time
5-15 early 25.9% on time
within 5 min 18.5% on time
5-15 late 10.2% on time
15-30 late 8.2% delayed
30-60 late 7.5% delayed
1-2 hr late 5.4% 1 hr+ late
2 hr+ late 3.7% 1 hr+ late

Bins are not equal-width in minutes; column widths are uniform for legibility. Shares sum to 100%.

Why Flights Run Late

II. Where the lost minutes go BTS data

When a flight is at least fifteen minutes late, the airline codes the reason into one of five federal buckets. The single biggest is the most quietly damning: late-arriving aircraft - the same jet, running behind from its last leg, dragging the delay downstream all day. Read as a share of total delay minutes.

  1. 01

    Late-arriving aircraft

    the cascade: a plane that lands late leaves late

  2. 02

    Air carrier

    crew, maintenance, bags - inside the airline's control

  3. 03

    National Aviation System

    congestion, ATC flow, airport ops - absorbs most weather

  4. 04

    Extreme weather

    direct weather only; true weather footprint is far larger

  5. 05

    Security

Weather looks small here - under six percent. It isn't. A thunderstorm that jams an airport gets logged as National Aviation System, and the jet it stranded becomes someone else's late-arriving aircraft tomorrow. The weather bucket counts only the storm you were sitting in, never the ones upstream.

Most Punctual

III. Carriers by on-time arrival rate BTS data

They are closer together than the marketing suggests - the whole industry lives inside a fifteen-point band. But the order is stubborn: geography and network shape reliability. Hawaiian flies short hops in good weather; Delta runs a disciplined operation; the ultra-low-cost carriers, flying tight schedules with thin spare aircraft, sit at the bottom. Dots, not bars, because this axis starts at 65%.

Carriers, with cancellation rates
CarrierOn-timeCancelled
Republic (YX) 83.8% 3.2%
PSA (OH) 81.7% 1.2%
Endeavor (9E) 80.5% 3.3%
Delta (DL) 80.2% 0.2%
SkyWest (OO) 77.5% 0.9%
United (UA) 75.8% 1.2%
Alaska (AS) 74.5% 0.7%
Allegiant (G4) 74.5% 1.2%
Southwest (WN) 74.2% 0.3%
Spirit (NK) 72.8% 2.6%
JetBlue (B6) 72.0% 2.5%
Envoy (MQ) 71.7% 1.6%
American (AA) 67.8% 2.3%
Frontier (F9) 63.7% 3.5%

Worst Routes

IV. Where lateness is the schedule BTS data

Some city pairs are late almost every day, and it is rarely the airline's fault. They are short hops into the country's most crowded airspace - the New York and Chicago and Bay Area corridors - where a plane can be perfectly run and still sit on a taxiway waiting for a slot. On these routes, a third of arrivals miss the fifteen-minute line.

  1. 01

    PHL SJU

    Philadelphia to SJU
    46 min avg delay · ~4/day · see cause split
    44%
    on time
  2. 02

    ORD HSV

    Chicago O'Hare to HSV
    37 min avg delay · ~3/day · see cause split
    46%
    on time
  3. 03

    DFW ECP

    Dallas-Fort Worth to ECP
    29 min avg delay · ~5/day · see cause split
    47%
    on time
  4. 04

    PHL SEA

    Philadelphia to Seattle
    29 min avg delay · ~4/day · see cause split
    48%
    on time
  5. 05

    MDW MSP

    MDW to Minneapolis
    38 min avg delay · ~8/day · see cause split
    48%
    on time
  6. 06

    FSD MSP

    FSD to Minneapolis
    31 min avg delay · ~4/day · see cause split
    48%
    on time
  7. 07

    SJU BDL

    SJU to BDL
    51 min avg delay · ~3/day · see cause split
    49%
    on time
  8. 08

    MIA TPA

    Miami to TPA
    56 min avg delay · ~6/day · see cause split
    50%
    on time

Directional pairs; a route can behave differently in each direction. On-time = arrival within 15 minutes.

Schedule Padders

V. Buying on-time with the clock Illustrative v1

Here is the quiet trick behind rising punctuality: airlines fly no faster than they did fifteen years ago, but they now schedule more time for the same trip. Stretch the timetable and yesterday's late flight lands early today - no new engines required. The gap below is the block time added to the same city pairs since roughly 2010.

0h 1h 2h 3h 4h 5h 6h LAX - JFK Los Angeles to New York 5h29 5h48 +19 ORD - LGA Chicago to New York 2h19 2h35 +16 ATL - LAX Atlanta to Los Angeles 4h58 5h12 +14 SFO - ORD San Francisco to Chicago 4h20 4h33 +13 BOS - DCA Boston to Washington 1h29 1h39 +10
~2010 schedule · today's schedule · minutes added · avg +14 min across these pairs
Block times, in numbers
City pair~2010TodayAdded
LAX - JFK 5h29 5h48 +19 min
ORD - LGA 2h19 2h35 +16 min
ATL - LAX 4h58 5h12 +14 min
SFO - ORD 4h20 4h33 +13 min
BOS - DCA 1h29 1h39 +10 min

Hub Roulette

VI. Reliability of the connecting airports BTS data

A connection is only as good as the hub it runs through. Book the same itinerary through Atlanta or through Newark and you are making a genuinely different bet on whether your bag - and you - make the second flight. On-time departure rate for each major hub; the line marks the national average.

ATL Atlanta Delta
80%

SFO San Francisco United
78%

EWR Newark United
74%

DEN Denver United / Southwest
70%

CLT Charlotte American
70%

ORD Chicago O'Hare United / American
69%

IAH Houston United
68%

DFW Dallas-Fort Worth American
65%

national average, 75%. On-time = departure within 15 minutes of schedule.

Hall of Cancellations

VII. The days the system stopped BTS data

A delay is an inconvenience; a cancellation is a broken promise. The worst days in modern American aviation were rarely about a single storm - they were about brittle systems failing under load. These four are a matter of public record, from DOT penalties and federal investigations.

  1. 01

    The Southwest holiday meltdown

    December 21-29, 2022
    Southwest Airlines ~16,700 flights

    Cause Winter Storm Elliott, then an outdated crew-scheduling system that could not re-assign crews once the network broke.

    Outcome Roughly two million passengers stranded over the holidays. In December 2023 the DOT fined Southwest $140 million - the largest consumer-protection penalty in the department's history.

  2. 02

    The day the NOTAM system failed

    January 11, 2023
    Entire U.S. system ~1,300 flights, 11,000+ delays

    Cause A corrupted database file took down the FAA's NOTAM pilot-alert system overnight.

    Outcome The FAA ordered the first nationwide ground stop of departures since September 11, 2001. Traffic recovered within a day.

  3. 03

    Delta and the CrowdStrike outage

    July 19-24, 2024
    Delta Air Lines (worst hit) ~7,000 flights

    Cause A faulty CrowdStrike security update crashed Windows machines worldwide; Delta's crew-tracking system took days to recover.

    Outcome About 1.3 million Delta passengers affected over five days. The DOT opened an investigation; Delta put the cost near half a billion dollars.

  4. 04

    The day flying stopped

    September 11-13, 2001
    All U.S. civil aviation every flight, ~2.5 days

    Cause After the attacks, the FAA grounded all civilian aircraft and closed U.S. airspace.

    Outcome The only complete shutdown of American commercial aviation in its history. It remains the benchmark every later 'ground stop' is measured against.

Methodology

VIII. What is real, and what is not

The tagged sections below are now computed from the real BTS record for June 2024 (one month). Sections still marked illustrative have not been recomputed. Every figure on the page is tagged. Read the tags.

BTS data means a published, verifiable DOT/BTS figure, a documented public event, or a value computed directly from the loaded BTS record. Illustrative v1 means a representative stand-in - the number is plausible and in the right range, but it has not yet been recomputed from raw records. Real BTS monthly files have been loaded and npm run data has been run, so the computed sections carry the real tag; anything still tagged illustrative is a stand-in the loaded month does not cover. Nothing here is meant to be quoted as fact while it wears the illustrative tag.

Source

The record is the U.S. DOT Bureau of Transportation Statistics, Reporting Carrier On-Time Performance database - the mandatory filing every large U.S. airline makes for every domestic flight. It is public, free, and enormous: roughly 7.3 million flights a year, monthly files back to 1987, at https://transtats.bts.gov/Tables.asp?DB_ID=120 (bulk zips under transtats.bts.gov/PREZIP/). The period framing this snapshot: June 2024 (one month).

What "on time" means

The federal definition is blunt and is the one used throughout this page: a flight is on time if it arrives less than 15 minutes after the scheduled gate time. Fourteen minutes late is on time; sixteen is not. Cancelled and diverted flights are neither - they are counted separately. The national on-time rate here, 75%, and cancellation rate, 1.3%, are the measured figures for June 2024 (one month).

How delays are coded

Since June 2003, any arrival fifteen or more minutes late is attributed to one of five buckets - carrier, weather, NAS (national aviation system), security, late-arriving aircraft - reported as delay minutes. Those five categories and their definitions are real. The split shown in "Why Flights Run Late" is computed from the delay-cause minutes in the loaded record; exact shares move year to year, and, as noted there, the weather bucket structurally undercounts weather's true footprint.

Section by section

  • The Shape of a Delay BTS data - bin heights are the arrival-vs-schedule distribution of the loaded record. The shape (mode a few minutes early, a long right tail) is the record's own.
  • Why Flights Run Late BTS data - categories and definitions real; the split is computed from the loaded delay-cause minutes.
  • Most Punctual BTS data - on-time and cancellation rates computed per carrier from the loaded record.
  • Worst Routes BTS data - on-time rate and average delay computed per directional route from the loaded record.
  • Schedule Padders Illustrative v1 - the industry-wide creep in scheduled block time is a real, documented phenomenon; the specific minutes per pair are representative (this section needs a decade-apart baseline the single-month load does not provide).
  • Hub Roulette BTS data - on-time rate computed for flights out of each hub from the loaded record.
  • Hall of Cancellations BTS data - four documented events, drawn from DOT enforcement actions and federal records.

What you're not seeing

International flights (BTS covers domestic). Small regional carriers below the reporting threshold. The reason your specific flight was late, which the data rounds into a bucket. And, in this snapshot, any trend across time: this is a single month, so there is no seasonality, no year-over-year, no time-of-day curve yet. Loading the real file did for the tagged sections exactly what it is meant to: it replaced the representative numbers with the distribution the record actually holds for this window.


Snapshot dated 2026-07-04. Source attribution: U.S. Department of Transportation, Bureau of Transportation Statistics. This is an independent editorial reading, not affiliated with or endorsed by the DOT.