Pro Flight Search is an independent flight-disruption intelligence site. This page explains, in plain English, what we track, how each score is built, how often it refreshes, and where the numbers stop being certain.
01
What we track
Pro Flight Search is a flight-disruption intelligence site. We continuously track scheduled and live flight operations at 200+ of the world's busiest airports and turn the raw signal into plain-English scores: how delayed an airport is right now and how exposed it is to weather.
Everything on the public pages is computed from the same pipeline that we describe below. There is no editorial fudging of the numbers. The figures move because the underlying flights and weather move.
02
Airport Delay Index & Pulse Score
The Delay Index is an industry-style 0 to 5 score, averaged across an airport's arrivals and departures, that captures how far today's flights are drifting from their published schedules. 0 means the airport is running to schedule; 1 to 2 is minor slippage; 3+ means material disruption with longer average delays and a wider spread.
The Airport Pulse Score turns that into a friendlier 0 to 100 read. It is a weighted composite: 60% the share of today's tracked arrivals and departures running on time, plus 40% the inverted Delay Index. We label it Excellent (85+), Good (70 to 84), Moderate (55 to 69) or Disrupted (below 55) so two hubs can be compared at a glance.
The global pulse on the delays dashboard is the composite of these per-airport indices across every tracked airport, so a single number summarises how the world's network is running at this moment.
03
Weather-Delay Risk score
The weather-delay risk score (0 to 100, per airport) reads the aviation weather an airport is actually reporting and estimates how much punctuality pressure it implies. It combines current flight rules (VFR / MVFR / IFR / LIFR), visibility, ceiling, wind and gusts, the upcoming terminal forecast, pilot reports, AIR/SIGMET advisories, and short-range forecast-model guidance.
Each input is capped so that a stack of correlated readings can't dominate the result. We then cross-check the weather against rolling delay-window history for that exact airport after a short lag, so a clear-morning snapshot is never mixed into an afternoon storm. The "vs clear-weather baseline" deltas shown in minutes come from weather observations matched to later live delay windows.
04
Cancellation risk after a delay
The cancellation-risk curves use the one thing final-state datasets throw away: every successive airport-board observation. We count chronological changes in the board's revised or estimated departure clock, then calculate eventual cancellation probability after the flight reaches 1+, 2+, 3+, or 4+ posted-time pushes. The separate minute curve uses the same 20 to 40, 40 to 70, and 70+ minute bands as our Board Honesty analysis.
Still-open flights never enter the denominator. A flight counts only after the departure board records cancellation or terminal/actual-departure evidence, and the newest two local dates are held back. Each published cell carries its resolved-flight sample and 95% interval; below 100 resolved flights the probability is suppressed. Hub, airline, and local departure-time cohorts remain independent rather than being blended into a falsely precise personal prediction.
05
Does a delayed flight make up time?
The delay-recovery analysis pairs the same operated flight's actual departure and arrival board clocks using flight number, exact origin and destination, UTC service date, and the nearest valid scheduled arrival. Recovery is departure delay minus arrival delay: a positive number means the flight reached its destination closer to schedule than it left.
Published cohorts begin at 15 minutes of actual departure delay and end at six hours, then split by scheduled gate-to-gate length: under 2 hours, 2 to 4 hours, 4 to 8 hours, and 8+ hours. Every cell carries its sample and remains suppressed below 100 paired flights. This is deliberately described as gate-to-gate schedule recovery: taxi time, routing, congestion, and timetable padding are part of the result, so the number does not isolate cruise speed or prove why time was recovered.
06
Data signals
Our scores are built from a small number of independent signals:
- Live flight operations: scheduled times, on-time counts, cancellations and the delay index come from aviation-industry flight-tracking feeds.
- Official aviation weather: METAR observations and TAF forecasts, plus pilot reports and AIR/SIGMET advisories, sourced from official meteorological reporting.
- Open flight-event data: taxi times and operational milestones drawn from open performance datasets, used for the taxi-in metrics and network context.
- Historical patterns: day-of-week and time-of-day tendencies computed from a rolling window of our own per-airport snapshots.
- FAA NAS programs: ground-stop and ground-delay programs from the public FAA NAS status feed, shown as a cause line on US airport delay pages. We do not republish FAA's generic delay minutes (our boards already measure those) or NOTAM closures (those are GA restrictions, not passenger shutdowns).
- Checkpoint waits (pilot): official airport sensor feeds at a small set of US hubs. A missing wait is omitted, never shown as zero. Overflow past the ropes is not in the sensor.
We aggregate and score these ourselves; the public pages report the resulting Pro Flight Search figures rather than any single supplier's raw feed.
07
How often it updates
| Surface | Refresh cadence |
|---|---|
| Global airport delays & pulse | Hourly |
| Per-airport delay & weather pages | Hourly |
| FAA NAS cause lines (US delay pages) | Every 30 minutes |
| TSA checkpoint waits (pilot airports) | Every 30 minutes; omitted if older than 90 minutes |
| Weather-delay risk map & scores | Hourly |
| Cancellation risk after pushes / delay | Checked hourly; rebuilt at least every 6 hours |
| Departure-to-arrival delay recovery | Checked hourly; rebuilt at least every 6 hours |
| Product changelog | When material features, data, or methodology change |
Every page timestamps its own freshness near the top. If a figure looks stale, the timestamp will say so.
08
Limitations
- Live travel data changes quickly. A morning reading and an evening reading at the same airport can differ materially. Treat every score as a snapshot of the moment it was generated.
- Scores are signals. The weather-delay risk score describes pressure on the network. It is never a deterministic prediction of any single flight.
- Smaller airports need more history. Rare weather regimes or low-volume airports will have noisier "vs baseline" numbers until more matched windows accumulate.
- We are independent travel intelligence. Pro Flight Search is an independent travel-intelligence site. We are not an airport, an airline, or an air-navigation provider, and our figures are not official operational data. Always confirm with the airline's own status board on the day of travel.
09
How we report
Every number on Pro Flight Search is computed deterministically from tracked airport boards, aviation weather, live position feeds, and our own rollups. No language model or generative model drafts scores, copy, or forecasts.
The Pro Flight Search Newsroom was retired on 16 August 2026. Former article URLs return HTTP 410. Site updates are published in our changelog.