Late aircraft, weather, or the airline: what actually causes US flight delays

Ask a plane full of delayed passengers what went wrong and most will say weather. The carriers' own filings say otherwise. Across 11.3 million flights between June 2025 to May 2026, extreme weather accounted for 8% of reported delay minutes. Late-arriving aircraft accounted for 34%.

That gap is the single most useful thing in the DOT's delay data, and it is almost never reported.

The five categories, and what they mean

Since 2003, carriers have been required to assign every delay minute to one of five causes. The definitions are narrower than the words suggest.

Late-arriving aircraft (34% of minutes nationally). The aircraft scheduled for your flight arrived late from its previous leg. This is not a cause so much as a symptom being passed along, and it is by far the largest bucket.

Carrier (27%). Anything inside the airline's control: crew scheduling, maintenance, fuelling, baggage loading, aircraft cleaning, catering, a captain timing out of legal duty hours. This is the category where good and bad operations genuinely separate.

National Airspace System (30%). Air traffic control volume, runway configuration changes, ground stops, airport construction, and non-extreme weather that reduces capacity without closing anything. Note that last part carefully.

Extreme weather (8%). Conditions severe enough to prevent operation: a blizzard, a hurricane, a thunderstorm cell sitting over the field. Only the genuinely unflyable counts here.

Security (0%). Terminal evacuations, screening equipment failures, a breach requiring re-screening. Rounds to nothing in most months.

Weather is bigger than 8% and smaller than you think

The extreme weather figure undercounts weather's real contribution, and the reason is the NAS category. When low ceilings at San Francisco cut the arrival rate from 60 an hour to 30, that is weather causing a delay, but it is filed under NAS because the airport is still operating. A large share of the NAS bucket is weather in disguise.

Add the two together and something like a third of all delay minutes have weather somewhere in the causal chain. That is a real number and worth knowing.

It is still not most of the delay. And more to the point, the weather share is the part no booking decision can change. What a booking decision can change is the exposure to the other two thirds.

Why "late aircraft" is the number that should change your behaviour

A delay charged to late-arriving aircraft is, by construction, a delay that happened earlier in the day to someone else. It travelled to you.

That makes it the only major category with a clean, actionable countermeasure: fly before the cascade starts. Nationally, the 5am–8am departure window runs 81.3% on time against 69.5% for 5pm–8pm. The aircraft on that early departure has no inbound leg to inherit from.

It also explains why on-time rates degrade through the day in a smooth curve rather than jumping around. Cascades compound. One 12-minute delay at 7am becomes 20 minutes by noon and 45 by evening, because each turn has less slack than the delay it is trying to absorb.

Where carrier-controlled delay actually shows up

The carrier bucket is the one that separates operations flying identical schedules through identical airspace. Two airlines running the same route out of the same airport face the same NAS constraints and the same weather. If one arrives on time ten points more often, most of that difference lives in the carrier column and in how well it protects its schedule from cascades.

Concretely, carrier-controlled reliability comes down to a few unglamorous things: how much ground time the schedule allows for each turn, how many spare aircraft and reserve crews are positioned where problems happen, and how aggressively the airline cancels early to protect the rest of the day rather than letting one aircraft poison twelve downstream flights.

That last one is why cancellation rate and on-time rate have to be read together. An airline that never cancels can post a worse punctuality number than one that cancels decisively, and be the better operation.

Reading a cause breakdown on this site

Every route, airport and airline page carries the five-way split for that entity. Three patterns are worth recognising:

Late-aircraft dominant (over about 35%). The operation is running with too little slack. Book the first flight of the day, and treat afternoon departures as materially riskier.

NAS dominant (over about 32%). The constraint is the airspace or the airport, not the airline. Changing carriers will not help much. Changing airports might: a secondary field in the same city often has a completely different NAS profile.

Carrier dominant (over about 35%). This is a management outcome. The comparison table on the same page is where you find out who is doing better on the identical route.

One caution about the split itself. It covers reported delay minutes, not flights, so a small number of very long delays weighs as much as a large number of short ones. And cause reporting is only required from carriers above a revenue threshold, so the thinnest operations in the data contribute no breakdown at all.

What the five categories leave out

One gap is worth naming: the reported minutes describe delay that was charged to a cause, not delay that a passenger experienced.

A flight that departs on schedule into a headwind and lands twenty minutes late generates a late arrival with no reported cause at all, because nothing went wrong that anyone is accountable for. A flight held on the taxiway for an hour and then cancelled contributes no arrival delay whatsoever, because it never arrived.

So the cause split answers a narrower question than it appears to: of the delay minutes that carriers attributed to something, where did they go. It is still the best available view of why an operation runs late, and it is the only one published under a common definition across every US carrier. But it is a view of the reported subset, not of everything that made you late.

---

See the breakdown for a specific operation: Denver (DEN) delay causes · United's on-time record · How the 15-minute rule works