Flight Arrival Buffers: What 54 Months of U.S. Airline Data Show

What Does Deadline Buffer Mean?

Deadline buffer = latest acceptable airport arrival minus scheduled arrival. A flight scheduled for 2:00 PM has a 60-minute buffer before a 3:00 PM airport-arrival deadline. Airport arrival here means reaching the scheduled destination gate, not collecting baggage, leaving the airport or reaching an event.

For each historical operation, we checked whether it reached that gate no later than its scheduled arrival plus each buffer. A zero-minute buffer includes early arrivals and arrivals exactly on schedule. Every threshold is inclusive. The same operation can succeed at several thresholds.

How Often Did Flights Arrive Within Each Buffer?

The chart pools all 54 months by adding successful arrivals and known outcomes. Each operation has equal weight; these are not averages of monthly percentages. Cancellations and diversions that did not reach the scheduled destination remain in the known-outcome denominator as failures.

On a narrow screen, scroll the table horizontally. With a keyboard, focus the table region and use the arrow keys.

The observed rate is successful arrivals divided by 33,432,885 known outcomes. Bounds divide by all 33,432,889 identifiable scheduled operations: the lower bound treats all four unknown outcomes as failures; the upper bound treats all four as successes. Both bounds round to the displayed rate at two decimal places for every buffer. The downloadable aggregate source preserves their full precision. These bounds are not confidence intervals and do not measure uncertainty about future flights.

How Were Cancellations, Diversions and Unknowns Counted?

• Ordinary arrivals: use the reported arrival delay, ArrDelay. A delay no greater than the buffer is a success.
• Cancellations: all 591,008 are known failures at every buffer shown. Passenger rebooking is outside this operation-level measure.
• Diversions: 85,090 operations were diverted. Those reaching the scheduled destination use DivArrDelay; the 8,310 that did not reach it are known failures.
• Unknown outcomes: four operations lack a resolvable outcome, including one diverted operation. They are excluded from the main rate and included in the bounds. A destination-reaching diversion can still be unknown if its arrival delay is missing.

The inclusive 15-minute result counts an ordinary arrival exactly 15 minutes late as a success and can include qualifying destination-reaching diversions. BTS's conventional on-time definition requires arrival less than 15 minutes after schedule. Its ordinary on-time classification and denominator differ from this custom calculation; see the BTS On-Time Reporting Directive. These results are not official BTS on-time percentages.

Methods and Sources

Source: the U.S. Bureau of Transportation Statistics Marketing Carrier On-Time Performance dataset, covering January 2022 through June 2026. It reports domestic nonstop operations for the reporting airline networks, including regional code-share partners. BTS's domestic reporting definition includes U.S. territories. This reporting frame is not every U.S. flight and can change over time.

Counts and exclusions: 33,432,890 source records, with one documented August 2024 exclusion because a missing operating flight number prevented safe physical identification. That leaves 33,432,889 identifiable scheduled operations, 33,432,885 known outcomes and four unknowns. The physical key uses date, operating airline and flight number, origin, destination and scheduled departure. Repeated physical keys are rejected; marketing aliases are not treated as additional independent flights.

Verification: all 54 monthly quality and curve outputs were checked for complete months, the seven thresholds, exclusions, consistent denominators, diversion counts and nondecreasing success counts. Monthly integer totals reconcile exactly to the pooled results. This publication reuses those audited outputs; it does not claim a fresh read of every raw record. The downloadable aggregate source contains the seven rows, unrounded fractions, bounds, controls and source provenance.

Scope: these are full-period observed operation counts, not sampled training rows. They do not use or validate MakeItBy's frozen predictive model, and its schedule-reconstruction filters are not this article's exclusions. See MakeItBy's historical prediction testing for that separate evidence. BTS's reporting instructions explain gate arrival, cancellations and diversions.

What Do These Results Mean for Flight Planning?

Later thresholds include more of the same historical arrivals. That arithmetic does not establish a causal benefit from changing a passenger's plans, or show that a specific flight has the nationwide rate. Pooling does not adjust for route, airline, season, weekday or scheduled-arrival time. The incomplete 2026 calendar year and unequal monthly flight counts also affect the mix.

Weather and network disruptions can affect many flights together. Millions of operations are not independent trials proving future reliability. These results do not measure missed connections, passenger reaccommodation, baggage collection, ground transport or event admission. Neither a larger buffer nor any percentage guarantees arrival before your deadline.

For a specific trip, choose your latest acceptable airport arrival and compare the flights you are actually considering. MakeItBy supports two to eight U.S. domestic nonstop flights on the same route and shows their scheduled arrivals alongside estimated deadline reliability. Compare Your Flights by Arrival Deadline. For a cruise, first calculate the airport-arrival deadline from your boarding requirements.

Related research: July vs. October flight reliability examines a separate seasonal comparison.