What a cancellation rate actually tells you (and what it hides)

A cancelled flight has no arrival time. It therefore appears in no on-time percentage, anywhere, ever. It is not counted as late. It is simply not counted.

Nationally, 1.80% of scheduled US domestic flights are cancelled. That is a small number that carries most of the worst outcomes in air travel, and leaving it out of a reliability judgement produces exactly the wrong ranking.

The incentive this creates

An airline facing a bad morning has a choice. Fly everything and deliver a day of four-hour delays, or cancel twenty flights early and run the rest close to schedule.

The second option produces a better published on-time rate. It removes the flights most likely to arrive late from the calculation entirely.

That does not make it the wrong decision. Cancelling early is often better for passengers than a rolling six-hour delay, because it triggers rebooking while seats still exist rather than at 11pm when they do not. But it does mean a punctuality figure read on its own can rank a cancellation-heavy operation above a more determined one.

This is why the grade on every page of this site is built from two numbers rather than one. The letter comes from the on-time rate. The plus or minus comes from the cancellation rate: under 1% earns a plus, over 3% earns a minus. The exact thresholds are stated verbatim on the methodology page.

What the rate does not distinguish

The DOT records that a flight was cancelled. It records nothing about what happened next, and the difference between those outcomes is enormous.

A cancellation on a route flown six times a day usually means a three-hour wait. A cancellation on a route flown four times a week can mean two days, or a drive.

A cancellation notified 14 hours ahead lets you rebook from home. One announced at the gate at 9pm means finding a hotel.

A cancellation with an available seat on a partner carrier is an inconvenience. One in the last week of December, when every seat in the market is sold, can end the trip.

None of that is in the data. The rate tells you how often it happens, not what it costs when it does. Frequency of service is the missing variable, and it is worth checking manually for any route you care about.

Why weather cancellations are different

Cancellation rates spike in winter, and the spike is concentrated. Most January days are normal; a handful shut down whole regions.

That has two consequences for reading the number.

An annual cancellation rate for a northern airport averages together 350 ordinary days and 15 severe ones. The average describes neither. Use the monthly strip on each page rather than the headline figure if your travel dates are fixed.

And weather cancellations are not evenly distributed across carriers at the same airport. When a hub carrier cancels, it cancels at scale and rebuilds the operation deliberately. A carrier with three gates and no spare aircraft cancels because it has no alternative. The rate can be identical; the recovery is not.

What a good number looks like

Under 1% is genuinely good and rare across a full year including a winter.

1% to 2% is normal for a large US operation.

Over 3% signals something structural: a thin schedule with no recovery capacity, a heavily weather-exposed network, or an operation running with too little margin. It earns a minus on this site's grade for that reason.

Diversions are worth a glance too, though they are much rarer. A diverted flight also has no arrival at its intended airport, and this site reports the rate separately on every page.

How to use it when booking

Read the two numbers together, always. An operation at 82% on time with 3.5% cancellations is not obviously better than one at 78% with 0.7%, and for a trip with a hard deadline it is clearly worse.

Weight cancellations more heavily when the trip has no slack, when the route is flown infrequently, when you are travelling in late December or during winter storm season, and when you are on a separate ticket with no protection.

Weight them less when there are five more departures the same day and nothing waiting for you at the other end.

The rate is on every route, airport and airline page on this site, next to the punctuality figure rather than buried below it, because reading one without the other is how people pick the wrong flight.

Checking frequency yourself, in two minutes

Since the cancellation rate cannot tell you what a cancellation costs, and frequency of service can, it is worth checking that by hand before any trip that matters.

Open any booking site and search the route for a single midweek date. Count the non-stop departures on your carrier. That number is your recovery capacity.

Six or more, and a cancellation is an inconvenience: there is another aircraft in a few hours and probably empty seats on it. Three, and you are likely looking at a same-day recovery only if you are early in the queue. One or two, and a cancellation means tomorrow, or a different airport, or a drive.

Then check the same route on a competing carrier. Interline rebooking has narrowed sharply across the industry, but a route where two airlines each fly it four times a day is structurally safer than one where a single airline flies it eight times, because a problem that grounds one operation rarely grounds both.

None of this is in the DOT data. All of it takes two minutes and changes the decision more often than a point or two of published on-time rate does.

---

See both numbers together: Frontier's on-time record · Alaska's on-time record · Minneapolis (MSP) delay statistics