The Least Reliable Hour of the Week

Summary

Flights scheduled 6–9 PM on Sunday arrived on time 63.8% of the time, the lowest of the 42 combinations of weekday and time of day at the covered airports, against 89.8% for Wednesday 5–9 AM. Sunday was the worst weekday evening at 44 of 60 airports with enough flights to compare. The gap between the best and worst weekday grows through the day, from 2.0 points at 5–9 AM to 10.8 at 6–9 PM. Time of day still matters more than the day: at a typical airport the hourly averages sit 7.9 points from the overall rate and the weekday averages 2.2.

Every flight is placed by the day of the week and the hour it was scheduled to leave, in the origin airport’s local time. Each figure is the share of those flights that arrived fewer than 15 minutes late. The six windows are the ones used across the site. Weekdays follow the scheduled departure date, so a 1 AM departure on a Sunday belongs to Sunday. The network figures sum the departures of the 64 covered airports, the ones FlyerIntel profiles.

Every rate is a bucket with at least 120 flights, built from grid cells that each cleared the same floor. The 9 PM–5 AM window includes the small hours of the same calendar day.

The weekdays are close together in the morning and far apart in the evening. At 5–9 AM the best and worst weekday differ by 2.0 points. By 6–9 PM the same comparison is 10.8 points. Sunday starts about where other days do, 87.8% on time in the morning, and ends 24.0 points lower by evening. Tuesday ends its day 14.9 points lower than it began. Cancellations move the same way: 2.78% of Sunday 6–9 PM departures against 1.72% on Tuesday.

Weekday evenings are 6–9 PM departures. An airport appears only where all seven weekday evenings had at least 500 flights, which leaves 60 of 64 covered airports. The morning gap is the same comparison for 5–9 AM. Ranked by the evening gap. The lowest of seven weekday figures is partly the noisiest of them, so a single airport’s worst day is less firm than the count of airports that share it.

Across the 60 airports, the lowest weekday evening is Sunday at 44, Thursday at 11, Friday at 4 and Monday at 1. Where it is Sunday, 29 of the 44 are below the next-lowest day by more than the flight counts would produce by chance. The heatmap on each airport page names a single least reliable slot, an hour on one day, for 61 of 64 airports. Of those, 47 are on a Sunday, 10 on a Thursday, 3 on a Friday and 1 on a Saturday. The same page compares how far each hour’s average sits from the airport’s overall rate with how far each weekday’s does. For a typical airport the hours are 7.9 points away and the weekdays 2.2. Hour is the larger at all 64 airports.

BTS codes a cause only for flights that arrived 15 or more minutes late, so these are shares of the minutes on late flights. Airlines choose the code; the table describes how they filed, not a finding about fault.

Sunday evening is not a different sort of trouble. The split among the five cause codes is close to what the other six evenings show, with late-arriving aircraft the largest at 49.4% against 48.3%. What differs is the amount: 27.0 coded minutes of delay per completed flight against 21.8. That fits a system running fuller and recovering less, which is an inference from the mix and not something the codes state.

The grid has no date, so it cannot say how much of a weekday’s figure comes from a few unusual days. Holiday weekends fall on particular weekdays, and Sunday evenings after long weekends are in the Sunday figure, as are storms. The grid also has no measure of how busy each slot is, so it cannot say whether a slot is worse because it is fuller. It covers only the airports FlyerIntel profiles. The windows hide the hour-by-hour shape, which the airport pages show. And a rate for a weekday and window is an average over every one of that weekday in the source period, not a forecast for the next one.