Not much to add here conceptually. Here’s a few arcane details & insight into the thinking …
My company’s policy is to normally stop boarding about 7 minutes prior to published departure time. If you’re not past the boarding pass scanner & terminal-to-jetway door by then, you’re not getting on. Getting everybody seated, bags stowed, bins closed, seatbelts checked, PAs made, and gate-checked strollers & wheelchairs stowed below decks takes that last 7 minutes. Then we push back as soon as we can, typically on-time +/- 1 minute.
The computers are tracking all the passengers. Once the gate agent gets notice that all expected passengers are on board, they *will *close the boarding door even earlier. Like up to 15 minutes prior to departure. Which leads to us actually pushing back about 8 early.
So if you are running to try to make an earlier flight than your reservation says, you may well be left behind. Conversely, if you’re running to make the flight in your reservation, we know that too and will normally wait at least a couple minutes for one person or several minutes for several people. The agent knows who is running from which gates and when their previous flights arrived. So they know about when to expect the runners to get to our gate. I don’t know the specific decision logic they (or the big computer in the sky) use, but waiting a 3-5 minutes for runners is common.
And that’s the story for a normal run-of-the-mill departure. Then come the special cases.
Our real goal is to *arrive *on time or early, not necessarily *depart *on time. DOT on-time statistics are based on arrival time. If we’re heading into a hub or an international gateway, folks depend on us arriving on time to make their connections. So the company plans, and we execute, to make that goal.
The published flight schedule starts with an expected taxi-out time based on the specific airport & time of day & planned congestion. Same for taxi-in. The scheduled time from takeoff to landing is based on the cruise speed of the planned aircraft type, the statistically typical route, the statistically typical winds for that time of year, and a factor for expected airborne congestion. Then some padding is added to cover for the daily “stuff happens”. The published arrival and departure times you see are built from that total duration.
If weather, airport or ATC issues, unusual winds, or a change in aircraft type mean that today’s flight will run longer than scheduled, we try to solve that. The best (read “cheapest”) way to do that is to depart early. Which we will do up to 15 minutes, but not at the cost of deliberately leaving people behind. If that’s not enough or we can’t leave early and still accomodate everybody, then we leave as soon as we and we can fly up to ~10% faster. At enough marginal extra cost to guarantee this flight was unprofitable.
Finally, occasionally we get into the “here comes a big storm; we either leave 30 minutes early or 3 hours late.” Or the equivalent situation is expected at the destination at our arrival time. That’s pretty darn rare though. Rare as in I do that once every 2 or 3 years of full-time flying. But if we are in that scenario, then we’ll leave early with as much load as has shown up. Sucks to be you if you get there after we’re gone.
A helpful tip: If you check in online & get your boarding pass before you leave for the airport, we *know *to expect you. If you plan to check in when you get to the airport, we *don’t know *to expect you until then. So if you’re delayed in traffic, you may get left behind, whereas if you already had a boarding pass we might not have left early or might even have waited a few minutes.