Flight reliability check before you book a business trip

Build me a flight reliability lookup that my assistant and I open before booking any business trip. The problem it solves: the cheapest itinerary is often the one that strands you, and normal booking tools show price and schedule but never on-time performance. This app puts reliability in front of the booking decision. Use AeroDataBox for all flight and airport data and Notion for saving the final choice.

Main search screen. I enter an origin and a destination using airport autocomplete, powered by the AeroDataBox Search Airports by Free Text operation, so I can type a city or airport name and pick the right airport without knowing codes. I then pick a travel date. Optionally I set a must arrive by time and a time of day filter for departures. Remember my recent and favourite airport pairs so repeat lookups are one click.

Building the flight list. First confirm the route is actually served and get a sense of daily frequency using Get Airport Daily Routes Statistics for the origin airport, checking that the destination appears in its route distribution. Then resolve the concrete flights operating that day using Get Airport Flights (FIDS) at the origin, filtering departures whose destination matches the chosen arrival airport. Important constraint: FIDS time windows are capped at twelve hours, so to cover a full travel day either issue two consecutive windows and merge the results, or narrow the window using the time of day filter I set. Airport codes accept IATA or ICAO, and dates are local airport dates, so pass the date exactly as the local date at the origin airport.

Scoring reliability. AeroDataBox delay statistics are keyed on a specific flight number such as BA286, so the app must resolve the route to concrete flight numbers first via the steps above, then call Get Flight Delay Statistics once per candidate flight. For each flight show the median arrival delay, the share of arrivals more than thirty minutes late, and the cancellation rate. Because each candidate costs a separate call against a RapidAPI quota, cap or page the results list: fetch statistics for a sensible first batch of flights (roughly ten) and let me load more on demand rather than fanning out across every flight on the route. Cache statistics per flight number for the session so re-sorting and comparing never refetches.

The must arrive by rule. When I set a must arrive by time, grey out (do not hide) any flight whose scheduled arrival plus its typical delay would realistically blow the meeting, and label why, for example "typically lands 14 minutes after your cutoff". Keep greyed flights selectable so I can override the recommendation deliberately. Let me tune the buffer, since a client meeting deserves more slack than an internal one.

Compare view. Let me shortlist several flights and open a side by side compare, showing departure and arrival times, median delay, share late over thirty minutes, cancellation rate, and the resulting verdict in one table so the trade off between an early departure and a more reliable one is obvious at a glance.

Current conditions. Show a live status strip for both the origin and destination airports using Get Airport Delay Snapshot, so I can tell whether today is already going badly at either end before I commit to an itinerary routing through it.

Saving the decision. When I choose a winner, save it to my Notion booking log using Create a Page, capturing the route and date, the flight number and airline, scheduled departure and arrival, the three reliability numbers, the current airport conditions at the time I decided, and a free text reason for why I picked it. Let me confirm or edit the reason before it saves. AeroDataBox is read only, so Notion is the only write in this app.

Make the ranking explainable everywhere in the UI. The entire value of this app is defending the booking choice to whoever is paying for it, so never show a bare score. Always show the underlying numbers and a plain sentence explaining why one flight outranks another, for example "lands on time more often and is cancelled a third as much as the cheaper option an hour earlier". Handle the case where a flight has too little history for meaningful statistics by labelling it as unrated rather than silently scoring it well.