Flight Layover Time Calculator

Introduction: a layover is a probability, not a sum

The usual way to plan a connection is to take the airport's minimum connection time, add however long you think your inbound flight will be late, and call the total a safe layover. That arithmetic has a hidden and fairly serious flaw: it treats delay as a fixed number when it is a distribution. Adding the average delay to the minimum connection time does not give you a safe connection. It gives you, roughly, a coin flip — because about half the time the delay is worse than average.

What you actually want is a confidence level. "How long a layover do I need to make this connection nine times out of ten?" is a question with an answer, and it is a different answer from "MCT plus the average delay". This calculator asks for the arriving flight's on-time performance, which airlines and route-comparison sites publish, fits a delay distribution to it, and reports both the probability of making a given layover and the layover you would need for a chosen confidence.

The second thing it fixes is that the minimum connection time published for an airport is rarely the minimum connection time for your connection. The published figure assumes a straightforward same-terminal transfer. Change terminals, clear immigration, collect and re-check a bag, or fly on two separate tickets, and the real requirement climbs — sometimes by more than the published MCT itself. Those adjustments are inputs here rather than a paragraph of advice.

How to use the flight layover time calculator

1. Enter the published minimum connection time. Get it from the airline or the airport, not from a guess: it is airline-specific and it varies by terminal pairing, so the number for a domestic-to-domestic transfer at your hub is not the number for a domestic-to-international one.
2. Tick what actually applies to your connection. A terminal change, re-clearing security, immigration and customs, re-checking a bag, and travelling on separate tickets each add time, and they add on top of the published MCT rather than being included in it.
3. Enter the arriving flight's on-time performance. This is the percentage of that flight's arrivals within 15 minutes of schedule, which is the US Department of Transportation's definition and the figure booking sites display. If you cannot find the flight, the route or the carrier's overall figure is a reasonable substitute.
4. Pick a confidence level. Ninety percent is a sensible default for a trip that matters; ninety-nine percent is what you want when the connection is the last flight of the day or the ticket is non-refundable.
5. Enter the layover you have actually been offered to get the probability that it works, which is more useful than a yes-or-no verdict when you are comparing two itineraries an hour apart in price.

The formula: fitting a delay distribution to a published statistic

Let be the effective minimum connection time after adjustments, the scheduled layover, and the arrival delay of the inbound flight. You make the connection when the delay does not eat the slack:

To turn that into a probability we need a distribution for , and we have exactly one published number to fit it to: the on-time performance , defined as the fraction of arrivals within 15 minutes. Fitting a one-parameter exponential gives:

from which both answers follow directly. The probability that a given layover works, and the layover required for a chosen confidence , are:

and the effective minimum connection time is the published figure plus whatever your particular transfer requires:

Plain-text formula: effectiveMct = publishedMct + sum(adjustments); lambda = -ln(1 - otp) / 15; probability = 1 - exp(-lambda * (layover - effectiveMct)); requiredLayover = effectiveMct + 15 * ln(1 - confidence) / ln(1 - otp), with every time in minutes and on-time performance and confidence as fractions.

The ratio in the second expression is worth noting for its own sake: the slack you need is 15 minutes multiplied by the ratio of two logarithms, and it does not depend on the minimum connection time at all. A flight with 80% on-time performance needs 21 minutes of slack for 90% confidence whether you are connecting at a small airport with a 30 minute MCT or a hub with a 90 minute one. The MCT sets the floor; the on-time performance sets the cushion.

Worked example: a terminal change on an 80% flight

Your inbound arrives at a hub with a published 45 minute minimum connection time, but the onward flight leaves from a different terminal, which adds 15 minutes. There is no immigration and no bag to re-check, so the effective minimum connection time is 60 minutes. The inbound flight arrives within 15 minutes of schedule 80% of the time.

1. Fit the distribution: λ = −ln(1 − 0.80) ÷ 15 = 1.6094 ÷ 15 = 0.1073 per minute.
2. The airline offers a 90 minute layover, so the slack is 90 − 60 = 30 minutes. The probability of making it is 1 − e^{−0.1073 × 30} = 96.0%.
3. For 90% confidence you need 15 × ln(0.10) ÷ ln(0.20) = 15 × 1.4307 = 21.5 minutes of slack, so a layover of 82 minutes.
4. For 99% confidence the slack rises to 42.9 minutes and the layover to 103 minutes.

Now compare that with the additive approach. This flight's average delay, under the same fitted distribution, is 1/λ = 9.3 minutes, so "MCT plus the average delay" gives 60 + 9 = 69 minutes — which the model puts at only 63%. That is the whole problem with averaging: it produces a layover that fails well over a third of the time, while feeling prudent because a delay allowance was included.

The table shows how quickly a poor inbound punishes you. Dropping from 90% to 60% on-time performance nearly triples the slack needed for a 90% connection, from 15 minutes to 38. It is usually cheaper to pick a better inbound flight than to buy an extra hour of layover.

What the published minimum connection time does not include

An airport's MCT is a scheduling constraint that airlines use to decide which itineraries their reservation systems will sell. It assumes a normal transfer of the type it was set for. Several very common situations fall outside it, and each is an input on this page:

The separate-ticket case deserves emphasis beyond the minutes. On a single ticket, a misconnection caused by a delay is the airline's problem: they rebook you, and depending on jurisdiction they may owe you a hotel or compensation. On two tickets you have simply missed a flight you paid for, and the second airline's obligation is nil. The extra hour in the table is a rule of thumb for the timing risk; the financial risk is not something a layover can buy off.

Limitations and assumptions

The exponential fit understates the tail. Real delay distributions have a long right tail: mechanical problems, crew timing out, weather closing a hub, and cancellations that make the delay effectively infinite. A one-parameter exponential fitted to the 15-minute on-time statistic describes the body of the distribution reasonably and the extreme tail badly. Treat a 99% figure here as optimistic, not conservative.

Cancellations are not modelled at all. If the inbound is cancelled, no layover helps. On-time performance statistics normally exclude cancelled flights, so the probability here is conditional on the flight operating.

On-time performance is defined at 15 minutes, and only there. The DOT standard counts an arrival as on time if it is within 15 minutes of schedule. That single point is all the calculator has to fit a whole distribution to. Two flights with identical 80% figures can have very different delay profiles beyond that point.

Published minimum connection times vary more than this suggests. They differ by airline, by terminal pairing, by whether the connection is domestic-to-domestic or international, and by time of day at some airports. The adjustments in the table are representative, not authoritative.

Departure-side risk is ignored. The model assumes the onward flight leaves on schedule. In practice a delayed onward departure works in your favour, which makes the estimate conservative in that one respect.

Nothing here is an entitlement. Making the geometric connection is not the same as the airline holding the flight, and a comfortable probability is not a guarantee. Check the rebooking policy, prefer earlier departures so later ones remain as fallbacks, and avoid the last connection of the day where you can.

Common questions about layover planning

Sources and assumptions. On-time performance is used with the US Department of Transportation definition, an arrival within 15 minutes of the scheduled time; that 15 minute threshold is the single data point the delay distribution is fitted to. The distribution is a one-parameter exponential, chosen because it can be pinned exactly by one published statistic and because it is the standard memoryless model for waiting times; it describes ordinary delays reasonably and understates the extreme tail, and it excludes cancellations entirely, so all probabilities here are conditional on the inbound flight operating. Minimum connection times are set by each airline and airport and vary by terminal pairing and by whether the transfer is domestic or international, so the published figure must come from your airline; the airport values shown are illustrative only. The adjustments for a terminal change, re-clearing security, immigration and customs, bag re-check and separate tickets are representative planning allowances rather than published constants. This page computes probabilities from the figures you enter, stores nothing, and does not predict whether any particular flight will be late.