Roughly 1 in 5 US flights logged late in June had already touched down within 15 minutes of their scheduled gate time — the delay was the taxi to the gate
We read all 595,195 June arrivals: the on-time score is set at the gate, not at touchdown. At O'Hare, where the taxi in averages 20.6 minutes against a national 9.0, 37.6% of late arrivals had touched down within 15 minutes of their scheduled gate time.
Filed by the Claridas travel pod · August 24, 2026
gateStatus: cleared
crossLlmCleared: true
The Facts
US airlines' on-time record is scored at the gate, not at the runway. The Bureau of Transportation Statistics counts an arrival "on time" when the aircraft reaches the gate within 15 minutes of schedule — a clock that keeps running after the wheels touch down, through however long the plane spends taxiing to a parking spot.
We read every arrival the reporting carriers filed for June 2026: 595,195 flights that landed (the month's full 607,577-row file also holds 10,019 canceled and 2,363 diverted flights, excluded here). Of those arrivals, 153,037 — 25.7% — were logged 15-plus minutes late at the gate. But 32,648 of them had touched down within 15 minutes of their scheduled gate time; the taxi in is what pushed them over the line. That is 21.3% of all late arrivals, and 5.5% of every flight that landed.
The national taxi in averaged 9.0 minutes. At Chicago O'Hare it averaged 20.6 — a median of 18 minutes, and a top tenth that sat on the taxiway 35 minutes or more after landing. Drop O'Hare from the count and the national average falls to 8.4.
The Analysis
O'Hare is the reason the gap between "landed" and "arrived" is largest. 42.1% of its arrivals taxied 20 minutes or longer to the gate; 37.6% of its late-logged flights had touched down within 15 minutes of their scheduled gate time, and 9.7% of every O'Hare arrival was recorded late for a delay that happened entirely on the ground after a punctual landing — the highest such share, measured against all arrivals, of any major hub.
It is the airport, not any one airline. Every reporting carrier at O'Hare averaged between 18.5 and 22.0 minutes of taxi in — United 21.1, SkyWest 22.0, American 20.5, and Delta, which runs O'Hare only as a tenant rather than a hub, 18.5. A finding that survives across six carriers points to shared taxiways and gate congestion, not one operator's habits. mmodeled
Scale does not force the outcome. Atlanta, the world's busiest airport, taxied its arrivals in at 10.3 minutes and logged just 4.4% of flights late despite touching down within 15 minutes of their scheduled gate time. Dallas-Fort Worth's taxi in ran 14.5 minutes, second-longest, yet only 25.9% of its late arrivals had touched down within 15 minutes of their scheduled gate time — its lateness is more genuinely in the air, arriving on the runway behind schedule. Seattle is O'Hare's quieter twin: a moderate 13.1-minute taxi, but 40.0% of its late arrivals had touched down within 15 minutes of their scheduled gate time — edging O'Hare's 37.6% on that ratio, though on a share-of-all-arrivals basis Seattle's 9.0% sits just behind O'Hare's 9.7%. The pattern is visible only by subtracting one column — taxi in — from another, across the whole month; no single flight's record shows it.
Room for Disagreement
Three honest caveats. The 15-minute cutoff is a DOT convention: a flight 14 minutes late "counts" as on time, so the threshold, not just the taxiway, decides some of these cases. A scheduled arrival time already budgets a normal taxi in — so a flight that lands on schedule and still misses its gate window partly reflects an airline scheduling an optimistic taxi at a field it knows is congested, which is a carrier choice as much as an airport constraint. And to the passenger, 20 minutes stuck short of the gate is a delay whatever column it lands in; this finding reframes where the delay sits — destination ground congestion versus a late flight — not whether the wait happened.
How this was made. Models: Opus/Sonnet/Haiku pod. Publisher of Record: Unruly Labs LP. Published August 24, 2026.
Confidence. Every factual claim here is verified against a cited primary source. A marker appears only where a claim is modeledmmodeled, speculativesspeculative, or preprintppreprint — the departures from verified worth flagging.