Missed connections
A Socratic walk-through of missed connections — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why can a fifteen-minute delay on one flight cost some passengers an entire day?
One aircraft lands fifteen minutes late. Everyone on board absorbs the same fifteen minutes. Yet by evening most of those passengers are home only slightly later than planned, and one is asleep in an airport hotel, a day from where she meant to be. Nothing different happened to her. So where did the extra day come from, if not from the delay?
Reasoning it through
REASONING #The instinctive answer is that delays propagate, or that the airline handled it badly. Both accounts predict something we can check: if the cost came from the delay itself, it should be roughly proportional to the delay, and roughly the same for everyone on the aircraft. It is neither. Most passengers on that flight lost fifteen minutes exactly. One lost a day. Whatever produced the day is not a property of the delay at all.
So it must be a property of the passenger. What varies between them? Only one thing that matters: how much room each had between landing and their next departure.
Think about what that room is made of. Between wheels-on and a connecting door closing, a fixed sequence has to complete: taxiing to a stand, waiting for the jet bridge, walking a terminal or riding a transit train, sometimes clearing security or immigration, and arriving before boarding closes — typically ten to twenty minutes before the departure time printed on your ticket. Airlines publish this floor as a minimum connecting time, specific to the airport and often to the pair of terminals. What is left after it is the passenger's slack, and slack is the only part of a connection a delay can eat.
Now the shape of the thing becomes clear. If the delay is smaller than your slack, your cost is zero — not small, zero. One minute past it, and you are not on the aircraft. The relationship between delay and cost is not a slope; it is a step, and every passenger stands at a different distance from the edge.
That explains why the cost is discontinuous. It does not yet explain why the step is so tall. Why not simply the next departure, an hour later?
Because what you need is not the next flight but the next empty seat, and airlines run close to full — load factors in recent years have sat in the mid-eighties as a share of seats sold. On a busy route with hourly departures you may lose two hours; on a thin route with three departures a day you may lose most of them. And if the onward leg is a long-haul operating once daily, the step is exactly twenty-four hours by construction, with no intermediate outcome available. The height of the step is set by the timetable's granularity, not by the size of the delay.
Now the part that inverts the folk story. You might assume a better-run hub leaves fewer people near the edge. The opposite is true. Hubs schedule arrivals and departures in banks — a cluster of inbound flights, then a cluster of outbound ones — precisely so many connections can be made in a short window. That design minimises average connecting time, and minimising connecting time is the same operation as minimising slack. The tighter the bank, the larger the fraction of passengers sitting within a few minutes of a cliff. Fragility here is not sloppiness; it is the direct cost of the efficiency that makes the connection attractive at all.
One further compounding effect: a late arrival into a bank dumps everyone who missed onto the rebooking desk at once, competing for the same scarce seats, so the queue that decides your fate is itself a burst.
The analogy
THE ANALOGY #Think of catching the last train home. Leave the party at any time before a certain minute and lateness costs nothing at all; leave one minute after it and the cost is a night in a hotel or an expensive taxi. Nobody experiences a gradual worsening — the penalty is stored up at one instant, and where that instant falls depends on the timetable rather than on you.
a train you miss was never aware of you, whereas an airline sometimes knows you are running and can hold the departure — so the edge is partly negotiable, and this is the weakest joint in the account, since whether a flight waits depends on how many connecting passengers there are and what holding costs the airline downstream.
Clarifying the model
THE MODEL #Three refinements are worth keeping.
The step is a threshold, not a fault. Nobody made a mistake and no rule was broken; the outcome was fixed by arithmetic when the schedule was published months earlier.
Slack is not the same thing as schedule padding. Padding is buffer added to one flight's block time to protect its own punctuality; slack is a property of a particular passenger's itinerary, and two people on the padded aircraft can have entirely different amounts of it.
Slack is bought and paid for. A passenger can choose a longer connection — usually with a longer total journey, sometimes a higher fare — and is therefore picking a position on the cliff, generally without being told that is the choice. Airlines do the same at network level, padding block times and buffering banks, trading punctuality against aircraft utilisation.
And the recovery machinery matters more than the delay. Because the loss is a step, the damage depends on whether a path back exists: a seat on another carrier, a rebooking policy, a hotel. Two passengers with identical slack and identical delays can end up a day apart depending on which of those existed — a part of the story that is institutional rather than physical, and varies by carrier and jurisdiction.
A picture of it
THE PICTURE #How to readthis is a packet-layout diagram repurposed to draw minutes rather than bits, so read the numbered scale as minutes of one scheduled connection laid end to end. Everything from "Taxi in" to "Boarding cut-off" is the fixed floor the airport imposes; it cannot be compressed by hurrying. Only the final field is yours, and a delay eats into it from the left, costing nothing until it runs out. The figures are one plausible sixty-minute connection rather than measured data — the lesson is the proportion, that most of an hour is spent before you have any margin at all.
What became clearer
WHAT CLEARED #The day was never in the delay. A connection is a fixed floor of unavoidable minutes plus a slack remainder; a delay is free until it exceeds that remainder and catastrophic the moment it does. The height of the step is set by how often the next seat comes around, which is why a once-daily onward flight turns fifteen minutes into a day. And the hub design that makes short connections possible is the same design that seats most passengers close to the edge.
Where to go next
ONWARD #- How airlines decide whether to hold a departing flight for connecting passengers.
Key terms
TERMS #| Term | What it means |
|---|---|
| Minimum connecting time | the shortest interval an airport and airline will sell between an arrival and an onward departure. |
| Slack | scheduled connection time minus that minimum; the only part a delay can consume without consequence. |
| Bank (or wave) | a clustered set of arrivals followed by departures at a hub, arranged so many connections work at once. |
Every term the collection defines is gathered in the glossary.