How to Read On-Time Performance Stats Without Being Fooled

“On-time” sounds like a fact. It’s actually a definition — a specific, generous one — and once you know how it’s built, the headline percentages airlines quote stop looking so impressive. The number isn’t lying to you, but it’s answering a narrower question than you think you’re asking. Here’s how to read on-time stats without being fooled by them.

What “on-time” officially means

In US government data, a flight counts as on-time if it arrives (or departs) within 15 minutes of its scheduled time. That’s the threshold the Bureau of Transportation Statistics uses, and it’s the one nearly every published on-time rate is built on. The consequence is immediate: a flight that pulls into the gate 14 minutes late counts as “on-time,” identical in the statistics to one that arrived early. So does 1 minute late, and so does 14 minutes and 59 seconds. The metric is binary — on the right side of a line, or not — and it tells you nothing about how a flight performs inside that 15-minute window.

Arrival punctuality isn’t departure punctuality

These are two different numbers, and they answer two different questions. Departureon-time tells you whether the plane left the gate roughly on schedule; arrival on-time tells you whether you actually got where you were going on time — which is what you care about. A flight can push back on time and still arrive late, or leave late and make it up in the air. When a stat is quoted without saying which one it is, assume the more flattering reading was chosen. For connections, the arrival number is the one that decides whether you make your next flight; see how much layover time you really need.

The network average hides almost everything

A carrier-wide on-time rate is an average across thousands of flights — every route, every hub, every hour of the day, blended into one figure. That average can be perfectly fine while the specific flight you’re booking is consistently terrible. Two things it routinely conceals:

• Route-level variation. The same airline can run a model operation on one city pair and a chronic mess on another — different aircraft, crews, gates, and weather exposure. The brand average tells you nothing about your route.
• Time-of-day variation. Punctuality decays as the day goes on, as delays snowball through each aircraft’s rotation. A 7am and a 7pm on the same route can have wildly different odds, yet both vanish into the same daily average. That’s the whole reason morning flights have fewer delays.

The fix is to read on-time performance route by route and flight by flight, not by brand. A carrier’s logo on the headline number is the least useful thing about it.

The cancellation trap

This is the one that catches people. Cancelled flights are usually excluded from the on-time rate entirely — a flight that never operated can’t be 15 minutes late, so it simply drops out of the denominator. That means a carrier can post a strong on-time rate because it cancelled the flights that would have been disastrously late, scrubbing its worst days from the average. A clean on-time figure can sit right on top of an ugly cancellation rate. Always read the two numbers together: on-time performance and the cancellation rate. A flight that runs 88% on-time but gets cancelled often is not a reliable flight.

Padded schedules

Airlines have learned to game the 15-minute line by lengthening their published flight times. If a route genuinely takes 90 minutes but the schedule says 120, the plane can dawdle and still “arrive early.” The on-time rate looks great; your day got longer. Padding isn’t fraud — it absorbs real-world variability — but it does mean a sky-high on-time rate sometimes reflects a generous schedule more than a sharp operation. If a block time looks long for the distance, that’s a clue.

Sample size and confidence

A route that flies twice a week gives you a tiny sample, and a couple of bad weather days can swing its rate by double digits. A twice-daily route over a full season is a far more trustworthy signal. Be skeptical of any percentage built on a handful of flights, and weight stable, high-frequency routes accordingly. It’s also why a single recent horror story shouldn’t override a long, steady track record — and why a long, steady track record of mediocrity shouldn’t be excused by one good month.

How to read it, in practice

1. Find the on-time rate for your route and your departure time, not the airline overall.
2. Confirm it’s the arrival number, since that’s what affects you.
3. Check the cancellation rate next to it — never instead of it.
4. Sanity-check the block time for signs of heavy padding.
5. Make sure there’s enough flight history to trust the figure.

Do that and the headline percentage becomes one input among several instead of the whole story. If you want the underlying figures, the US DOT publishes the raw on-time data through the Bureau of Transportation Statistics and a plain-English summary in the Air Travel Consumer Report. To turn all of this into a booking decision, see how to pick a reliable flight and whether some airlines really are more reliable.