MIA→EYW

The weak window here is the middle of the day, not the evening: 11am–2pm runs 78.3% on time against 95.3% for 8pm–5am. That shape usually means a departure bank rather than an accumulating delay, so the recovery later in the afternoon is real rather than noise. The 5pm–8pm window is the busiest, with 315 flights scheduled in it.

February is the weak month at 75.2% on time, against 93.8% in May. Winter delays cluster: a single storm system can account for most of a month's shortfall, so an average December hides both very good weeks and genuinely unflyable days. January was the busiest month here, with 150 flights.

Across the week, Tuesday is the strongest day at 90.1% on time and Saturday the weakest at 77.6%. 13 points is a smaller lever than departure hour, but it is free.

Extreme weather accounts for 14% of delay minutes, well above the single-digit share typical across the system. Weather delays are lumpy: they arrive as whole bad days rather than as a steady drag, so the monthly strip above is the more useful view of the risk. The national share is 6%, so this operation carries roughly 2.3 times the usual weather exposure. The split is drawn from 8,922 reported delay minutes.

41% of delay minutes trace back to a late-arriving aircraft, meaning the plane was already behind before it got here. That is a cascade, and cascades start small and grow through the day. Across 8,922 reported delay minutes here, that single category outweighs weather by 3 to one. It is the strongest argument for booking the first departure of the morning, when the aircraft has been sitting at the gate overnight with nothing to inherit.

Cancellations are rare at 0.40%, below the national 1.81% and low enough to earn the plus on the grade. Roughly one flight in 100 is called off; the risk here is arriving late, not not arriving.