The Horror Story

It is the nightmare scenario. You are flying home for Diwali. You have your boarding pass. You arrive at the gate. And then you hear the announcement: “We are looking for volunteers to give up their seats for a voucher.”

Why does this happen? Did their database fail? Did they double-book a seat by mistake? No. They bet against you, and they lost.

The Economics of “Perishable Inventory”

An empty seat is the most expensive thing on a plane. It consumes fuel but generates zero revenue.

To understand overbooking, you have to understand the nature of an airline seat. If Amazon doesn’t sell a Kindle today, they can sell it tomorrow. It is Durable Inventory. If IndiGo doesn’t sell Seat 4A today, that value vanishes the moment the doors close. It is Perishable Inventory.

The Algorithm: Probabilistic Inventory Management

Airlines have decades of data. They know that on a Tuesday morning flight from Mumbai to Delhi:

• 3 businessmen will miss the flight due to traffic.
• 2 people will get sick.
• 1 person will miss their connecting flight.

The historical “No-Show Rate” is, say, 5%. If the plane has 180 seats, and they sell exactly 180 tickets, they will likely fly with only 171 people. That is 9 wasted seats.

So, the Revenue Management System sets the “Authorized Capacity” to 189 (105%). They intentionally sell tickets that technically do not exist.

The Trade-Off: Cost of Spoilage vs. Cost of Spillage

Product Managers constantly balance two risks:

1. Spoilage (Empty Seats): You didn’t sell enough, and the inventory expired. (Revenue Loss).
2. Spillage (Bumping Passengers): You sold too much, and now you have to pay compensation/vouchers/hotel stays to the angry customer. (Operational Cost + Reputation Damage).

The airline calculates that the Cost of Spillage (paying off a few passengers occasionally) is significantly lower than the Cost of Spoilage (flying empty seats every single day).

The PM Lesson: The “Thin Provisioning” Model

This logic applies to almost every digital product.

• Broadband ISPs: If every household downloaded a 100GB file at the exact same second, the internet would crash. ISPs sell “100 Mbps” speeds knowing that most people only use 5 Mbps on average.
• Cloud Computing (AWS/Azure): They assign more “Virtual CPUs” to customers than they have physical CPUs in the data center. They bet on the fact that not all servers will peak at the same time.

The Takeaway: If you design your system for 100% concurrent usage, you are often over-engineering and under-monetizing. Smart Product Management involves looking at the data, understanding user behavior, and taking a calculated risk on capacity.

Don’t build for the theoretical maximum. Build for the statistical reality.