How the index is built

Data contract

The source unit is a movement at a UK reporting airport. A matched operated movement is not the same unit as a scheduled flight. Short-notice cancellation means the CAA cancellation field only. An origin/destination is the final point of the service and is not necessarily nonstop.

Denominator law

D = M + C + U: matched operated movements, short-notice cancellations, and unmatched actual movements. Source delay buckets and cancellation shares use D. Published known-operation outcomes use M + C; unmatched pressure remains a data-quality measure.

Reconstruction

Integer bucket counts are enumerated inside the source rounding interval, with the additional requirement that their sum equals M. Multiple feasible vectors produce a range. A cell that does not reconcile is quarantined rather than guessed.

Metrics

Disruption is late-over-fifteen-minute bucket counts plus short-notice cancellations, divided by known operations. The operated Delay Severity Index applies fixed increasing weights to the nine delay buckets among matched movements and normalises by the maximum weight. Three-hour-plus delay is kept separate from cancellation.

Average positive gate delay is published as reported by the CAA, with early flights counting as zero minutes. A bucket-midpoint model is retained only as a diagnostic because its estimate does not use the same definition as the CAA published average.

Measurement regime

Timing is based on gate times where available. In the fallback described in CAA notes, runway times plus taxi-time estimates are used when gate times are unavailable. This measurement regime can affect comparability and is not a measurement of passenger journey time.

Archive epochs

The main archive begins in 2018. The 2020–2021 pandemic regime remains visible in history charts but is excluded from seasonal comparisons and P10–P90 profiles.

Direction and domestic pairs

National totals count airport movements. A domestic pair is observed at both UK reporting airports and therefore contributes a movement at each side; it is not one scheduled flight count. Separate arrival (A) and departure (D) aggregates are available in the derived dataset.

Publication gates

Cells with fewer than twenty known operations are suppressed. Cells from twenty to forty-nine are descriptive and are not ranked. A month with no observed events is described as such; it is not presented as zero risk.

See the corrections and quarantine ledger for excluded rows and reconciliation notes.

Quick answers

What does this record cover?

Published CAA movements, with arrivals and departures kept separate.

Does this predict my flight?

No. Historical monthly aggregates cannot establish the outcome or circumstances of an individual flight.