OnTimeRoutes

Methodology & data sources

Current statistics cover August 2025 to July 2026 (12 months). The most recent month published by the Bureau of Transportation Statistics is July 2026.

Where the data comes from

All flight statistics come from the U.S. Department of Transportation’s Bureau of Transportation Statistics (BTS) Reporting Carrier On-Time Performance database. US airlines with at least 0.5% of domestic scheduled-passenger revenue must report every domestic nonstop flight they operate: scheduled and actual times, cancellations, diversions and the cause of each delay. This is the same data behind the DOT’s monthly Air Travel Consumer Report. It is a US government work and in the public domain.

Airport names and coordinates come from OurAirports (public domain).

How often it is updated

BTS publishes each month’s data roughly two months after the month ends. We check for new data daily and load it the day it appears, then recalculate every page. Each page states the months it covers.

Definitions

Reliability grade

Each route, airport and airline gets a letter grade from a score of on-time % − 2 × cancellation %: A ≥ 82, B ≥ 75, C ≥ 67, D ≥ 58, F below. Cancellations are weighted double because they usually cost travelers far more time than a delay.

Minimum sample sizes

To avoid misleading numbers from a handful of flights, we only rank, and only ask search engines to index, pages with enough flights: at least 30 flights per month for a route, 25 per month for one airline on a route, 150 departures per month for an airport and 60 flights per month to appear in national rankings. Smaller routes are still shown with a note.

How we use AI

Every number on OnTimeRoutes is calculated directly from the DOT data by our code. On some pages, a short summary paragraph is drafted by an AI model (Anthropic’s Claude). The model receives only that page’s computed figures and is instructed not to add anything else. Before a summary is published, software checks that every number it contains matches the page’s data; summaries that fail are discarded. Summaries are regenerated only when the underlying data changes, and pages without a verified summary show only the computed “Key facts”. The AI does not write rankings, recommendations or any other content.

Limitations