Skip to content
DelayAtlas

Guide

Connecting flights: how much buffer the data says you need

Nationally, 8.2% of arrivals land an hour or more late. That single figure, not the airline's minimum connection time, is what should set your layover.

By Kaan Tokalı4 min readFigures from June 2025 to May 2026
Passengers walking through a wide airport concourse between gates

Airlines publish a minimum connection time for every airport, typically 35 to 50 minutes for a domestic transfer. Booking engines will happily sell you an itinerary at exactly that number.

The DOT data says that is a bet, not a plan. Nationally, 22.6% of arrivals are at least 15 minutes late and 8.2% are an hour or more late. A 40-minute connection survives the first category and loses to the second every time.

The minimum connection time is not a recommendation

The published minimum is an operational floor: the time the airline believes it physically takes to walk between the furthest realistic gate pair, including a terminal change and a bus if the airport requires one. It assumes the inbound flight arrives on schedule.

It is not a probability statement. It contains no information about how often the inbound flight is late at that airport, in that month, on that carrier. That information exists, it is free, and it is what should actually set the layover.

A framework built on arrival distributions

The useful question is not "how long do I need to walk" but "what fraction of inbound arrivals leave me enough time".

Take the inbound flight's own record. If 80% of arrivals are within 15 minutes of schedule and 8% are more than an hour late, then a 45-minute connection works about 80% of the time, a 75-minute connection works about 92% of the time, and a two-hour connection is near certain barring a cancellation.

Each additional half hour of layover buys progressively less. Going from 45 to 75 minutes eliminates most of the risk. Going from two hours to three eliminates very little, because what is left is cancellations and multi-hour disruptions that no layover survives.

That is the shape to hold in your head: buffer has sharply diminishing returns, and the steep part of the curve is between about 45 and 90 minutes.

Practical minimums

These are judgements, built on the distributions above rather than on airline policy.

Same carrier, same terminal, off-peak, small airport: 45 minutes is defensible.

Same carrier, large hub, midday: 60 to 75 minutes. Hub airports are where gate-hold and taxi-in delays live, and the walk is longer than you think at Atlanta, Dallas/Fort Worth or Denver.

Terminal change, or a hub during the evening bank: 90 minutes. The evening is the weakest departure window nationally at 67.7% on time against 89.3% for 5am–8am, and a terminal change at Chicago O'Hare or Newark is a real walk.

Separate tickets: two hours minimum, and understand that no one owes you anything. If the first airline delivers you late, the second is under no obligation to rebook you and will usually charge full fare for the next seat. This is the single most expensive mistake in the data, and it is entirely avoidable.

Winter, or a connection through a weather-exposed hub: add 30 minutes to any of the above, or route around it.

Two structural choices that beat any buffer

Connect early in the day. A morning connection has the rest of the day behind it. If you miss it, there are three more departures. An 8pm connection has nothing behind it but a hotel. This is the same argument as the first-flight guide, applied to the second leg instead of the first.

Connect through the right hub. Salt Lake City, Minneapolis, Charlotte and Detroit consistently run cleaner operations than Newark, LaGuardia or San Francisco. If two itineraries cost the same and one routes through a structurally better hub, that is a free improvement. The airport pages on this site carry each hub's own on-time rate and cause breakdown.

What buffer cannot fix

A cancellation is not a delay. If the inbound flight is cancelled outright, a four-hour layover helps no more than a one-hour one, and the relevant number is the cancellation rate rather than the on-time rate. Nationally that runs 1.80%.

This is where the difference between a route flown six times a day and one flown four times a week becomes the whole story. On the first, a cancellation costs you three hours. On the second, it costs you a day.

Check both figures before booking a connection that matters. Every page on this site prints them side by side, because reading one without the other is how people end up sleeping at an airport.

Your bag has its own minimum connection time

Airlines publish a minimum connection time for passengers. Baggage systems have their own, and it is usually longer.

A passenger can run. A bag rides a belt through a sorting facility on a fixed schedule, and at a large hub that transfer can take forty minutes on its own. This is why a tight connection sometimes delivers you on time and your suitcase the next day: the airline met its published minimum for you and missed it for the bag.

Two consequences worth acting on. If the connection is under an hour at a big hub, carry on. And if you must check a bag on a tight itinerary, accept that the bag is on a slower clock than you are and pack accordingly, because a delayed bag is not a delayed flight and appears in none of the statistics on this site.


Check your connecting hub: Charlotte (CLT) delay statistics · Chicago O'Hare (ORD) delay statistics · United's on-time record

Every figure in this guide is recomputed at build time from the US DOT Bureau of Transportation Statistics on-time performance database, covering June 2025 to May 2026. Methodology · Data sources