Why Airlines Sell More Seats Than They HaveOverbooking Is MathsContent for Intellectualsmagazine.berelentless.in
19 October 2026 · 7 min read · 222 reads
Overbooking looks like a mistake but it is a calculation. How it works, what it costs in trust, and what any booking business can learn.
You have a ticket. You paid for it. And yet the airline sold the same seat to someone else on purpose. That sounds like fraud, or at least a bad mistake.
It is neither. It is a calculation, and the airline made it before you ever booked. I wanted to understand how it works, because the same logic shows up in hotels, clinics, restaurants and any business that takes bookings. It also has a real cost, and the best-known case shows how bad that cost can get.
Wikipedia's page on overselling defines it as selling more of a service than you have supply for. It is common in travel and hospitality, and it is intentional: the seller expects some buyers to cancel or not show up. The goal, the page says, is to use close to 100% of what you have, because an empty seat on a plane earns nothing and cannot be sold later.
For airlines, this means booking more passengers than seats to cover no-shows. If everyone turns up, there is an "oversale." Then the airline has to take someone off the flight. In the industry that is called "bumping."
I could not find a standard no-show rate in the sources I used, and I do not want to guess one. So let me show the logic with an illustration. The numbers below are made up.
Say a plane has 100 seats. Say, out of every 100 people who buy a ticket, about 5 do not show up. (Illustrative.) If the airline sells exactly 100 tickets, it flies with about 95 people and 5 empty seats. Those 5 seats are lost money.
If it sells 105 tickets, then on an average day 100 people show up and the plane is full. Some days, 3 extra people show up and the airline has to find 3 volunteers or pay to move them. Other days, 7 do not show up and the plane has 2 empty seats anyway.
The airline is comparing two costs: the cost of an empty seat, and the cost of bumping a passenger (a voucher, a hotel, a refund, plus any damage to its name). It sells extra tickets until the next extra ticket would cost more in bumping than it would earn from filling a seat. In real life it uses past data for each route and day, not a round number like 5.
The page notes that Delta Air Lines began using algorithms based on historical data to set its overbooking levels in 2011. It also describes Delta's system of asking passengers at check-in how large a voucher they would accept to take a later flight, and bumping the lowest bids first.
Because bumping is rare, we forget it happens. The Wikipedia page cites US Department of Transportation figures from the February 2019 Air Travel Consumer Report. In 2018, 351,904 passengers on 17 major US airlines were bumped, and 10,938 of those were bumped involuntarily. That means a bit over 3 in 100 of the bumped passengers did not choose it. (The figures cover flights starting in the US.) So most of the people
Rules exist for this. In the US, airlines must first ask for volunteers. In the EU, Regulation 261/2004 sets compensation when volunteers are not enough. The page only sums these up, and the details change by case and country, so check the current rules before you travel.
The case everyone remembers
The best-known failure was United Flight 3411 on April 9, 2017, at Chicago O'Hare. This is a summary of what the Wikipedia pages report. After the passengers boarded, United needed four seats for airline crew and asked for volunteers to give theirs up. It offered vouchers, reported as rising to $800. Not enough people came forward, so four passengers were picked by United's system. One of them refused. He said he was a doctor and had
I will not retell the whole story, since the details are in public reports and I have only read summaries. As United itself said later, this may not be a pure overbooking case: the seats were needed for crew. It is still the best-known example of what happens when a seller needs seats back from people who already paid. What stands out for me is the business lesson. Whatever the maths behind a booking system, the
The same page points to a few other responses. JetBlue, it says, historically avoided overbooking and began doing it again in 2017. And in April 2017 Delta, after severe weather caused more than 3,200 cancellations in five days, paid unusually high vouchers, including over $11,000 to one group of three passengers in a single weekend. Both are reminders that the costs of overbooking are not fixed. They depend on how you treat the people on the
The page lists several other industries. Hotels do a version of it too, sometimes moving a guest to another hotel when they are full. (The page gives no firm source for some of the hotel details, so I will not lean on them.) Internet providers also sell more capacity than they could deliver if every customer used everything at once. That works until most customers do exactly that at the same time.
The most striking example on the page is internet service. It describes one cable broadband design where about 500 subscribers share 38 megabits per second of capacity, while each one is sold a plan advertised at 7 megabits per second. Each subscriber is guaranteed only a tiny fraction of that, about 0.076 megabits per second, which the page presents as a ratio of 92 to 1. The page is careful to say that this is not
The page also says plane makers do it. Airbus and Boeing, it says, overbook their own order backlogs, and use algorithms to estimate how many orders will be delayed or cancelled. So even a company that sells aircraft costing millions plans for some of its customers to change their minds.
What links all these cases is the same question: what is the cost of being wrong in each direction? An empty seat costs the price of the ticket. A slow internet connection costs some patience. A person who is turned away from a hotel at midnight costs far more than a room. The right level of overselling is different in each.
The page also lists ways to avoid the problem: non-refundable bookings, no-show fees, asking people to reconfirm, or adding capacity when a flight or room is oversold.
If you take bookings, you already face some version of this. A clinic, a salon, a coaching studio, a restaurant with reservations: each one has empty slots when people do not show up.
You have a few choices. You can accept the empty slots as a cost. You can reduce no-shows with reminders or a small deposit. Or you can overbook a little. The last one is the riskiest, because your customers will feel it in a way they do not feel with a plane ticket. A person who has to wait because you double-booked them may not come back.
If you do it, I think three rules from the cases above make sense. Know your real no-show numbers from your own records, not from a guess. Decide in advance what you will do for the person who loses out, and make it generous. And never let your customers find out in front of other customers.
Overbooking taught me that a smart calculation is only half of a decision. The other half is how it feels to the person on the wrong side of it. I am trying to look at any "it works on average" rule and ask who pays when it does not work.
Try this week: If you take bookings, count your last 30 appointments or reservations and note how many did not show up. Then write down, in one line, what you would offer someone if you ever had to turn them away. If you cannot finish that line, you are not ready to overbook.
Earn Embers for what you read
Sign in, answer a few quick questions about this piece, and earn Embers you can spend in the app.
Comments
Sign up free, read the story and mark it finished to join the conversation.
Loading comments…