A
ATL Passenger
Hartsfield-Jackson · ATL
Home/On-Time Performance/Methodology
Methodology

How the On-Time Performance data is built

Source

Flight performance figures come from the US Department of Transportation’s Bureau of Transportation Statistics (BTS), Reporting Carrier On-Time Performance (1987–present) dataset, published as monthly bulk CSV files at transtats.bts.gov/PREZIP. This is a US government work; BTS's own data is in the public domain with no access key or login required.

Enplanement/rank figures come from the FAA’s Air Carrier Activity Information System, published at faa.gov (Calendar Year 2024 final commercial-service enplanements workbook, plus the Calendar Year 2025 preliminary workbook where noted).

Dataset date and cadence

Data as of May 2026 (the most recent monthly file BTS had published as of this page’s build date, 2026-07-25 — BTS releases each month’s file roughly 2 months in arrears). This page reflects a single-month snapshot; it is not a live feed and is not currently on an automated refresh cadence. No automatic-update claim is made anywhere on this page.

Pipeline

  1. Fetched the full national monthly CSV (611,735 flight records for May 2026) directly from BTS’s PREZIP bulk file — no filtering at the source, full month.
  2. Filtered to every row where the flight’s origin airport or destination airport code is ATL — 55,067 ATL-linked flight records.
  3. Grouped the ATL subset by IATA_CODE_Reporting_Airline (the carrier code BTS itself reports) and computed, per carrier: total ATL-linked flights, cancellation rate, on-time arrival/departure rate (BTS’s own definition: not delayed ≥15 minutes), average arrival delay across all flights with a reported arrival time, and the sum of BTS’s 5 delay-cause-attribution columns (Carrier / Weather / NAS / Security / Late Aircraft).
  4. No records were merged across carriers, gap-filled, deduplicated, or invented. Every number on every carrier page traces to a straightforward aggregate of that carrier’s real rows in the source file. A carrier with no matching existing ATL Passenger airline profile page is shown by its bare BTS/IATA code rather than a guessed-at full name.
  5. Full-reparse recount, asserted in the build script before publishing: the sum of every carrier’s displayed flight count equals the ATL-filtered file’s total row count exactly (55,067 = 55,067).

Fields rendered vs. withheld

Rendered: carrier code (and name, where verifiable against this site’s own existing airline pages), ATL-linked flight count, arrival/departure on-time percentage, cancellation percentage, average arrival delay, and the 5-cause delay-minute breakdown — all real, all directly aggregated from BTS’s own columns.

Withheld: individual flight numbers, tail numbers, and specific flight dates/times (BTS reports these per-flight, but publishing 55,067 individual flight rows is a per-record page count far beyond what a useful reference page needs — this is an aggregate/summary presentation by design, matching the “no per-record pages when records exceed a few thousand” guidance this site follows elsewhere); diversion counts are tracked internally but not separately displayed (very low volume, and DOT’s cause-attribution system does not attribute diversions the same way as delays).

Why some airlines aren’t here

BTS’s on-time-performance reporting mandate applies to certificated US domestic carriers, not foreign flag carriers. ATL Passenger’s international airline profile pages (Air Canada, Air France, British Airways, KLM, Korean Air, Lufthansa, Qatar Airways, Turkish Airlines) are real, existing editorial pages — but no US DOT on-time performance figures exist for them in this dataset, so none are shown or implied here.

Corrections

If you spot an error in how a number here was computed from the underlying BTS file, or the BTS/FAA source data itself has since been revised, contact us and we’ll review it.