# Dynamic pricingDynamic pricingAutomatically adjusting prices in real time based on demand, competition or user behaviour to optimise revenue, margin or conversion.View full definition → and demand forecasting for perishable inventory
A Friday evening flight from Chicago to New York has 150 seats. The airline has sold 158 tickets. This is not a mistake. It is math.
Airlines routinely sell more seats than the plane holds because they know, from years of data, that a predictable share of passengers will not show up. That empty seat, if it flies unfilled, is worth exactly zero the moment the doors close. So is a hotel room at midnight, a rental car on Tuesday, and a cruise cabin at sail. This is the defining problem of travel and hospitality: perishable inventory, a unit that loses all value at a fixed deadline.
Revenue management is the discipline of squeezing the most money out of that inventory before it expires. Let us build the machine piece by piece.
A retailer with unsold shoes can discount them next month. A travel operator cannot. Once the flight departs or the clock strikes midnight, unsold capacity is gone forever.
The goal is not a high load factor (the percentage of seats filled) or a high average fare. It is maximizing total revenue per departure. A 100 percent full plane sold entirely at bargain prices can earn less than a 90 percent full plane with a smart fare mix.
You cannot price what you cannot predict. Forecasting estimates how many people will want each fare on each future date.
Travel demand follows strong, learnable patterns:
The core trap is censored demand. If a fare class sells out three weeks before departure, your booking data shows demand stopped there. But real demand kept climbing; you just had nothing left to sell. Naive models under-forecast popular dates because they only see what was booked, not what was wanted. Good revenue systems correct for this "unconstraining" of demand.
For a solid non-technical primer, the Cornell Center for Hospitality Research publishes accessible papers on hotel revenue management.
Once you forecast demand, you segment customers so each pays close to their maximum willingness to pay. You cannot ask travelers their budget, so you build fare fences: rules that separate price-sensitive customers from price-insensitive ones.
Common fences:
The fence is the point. Two people on the same flight in the same cabin can pay very different prices because they cleared different hurdles.
Here is the foundational logic, often called Littlewood's rule, the seed of modern yield management.
Imagine two fare classes: a discount fare and a full fare. You have limited seats. The question: should you sell one more seat now at the discount price, or protect it for a possible full-fare customer later?
Protect the seat only if the expected value of holding it beats the sure discount revenue.
Sell the discount seat if:
Discount fare > Full fare × P(full-fare demand exceeds seats protected)Rearranged, you keep protecting seats for full fare as long as:
P(full-fare demand > protected seats) > Discount fare / Full fareA quick example. Full fare is 400 dollars, discount fare is 100 dollars. The ratio is 0.25. So you keep protecting seats for full-fare travelers as long as the probability of selling them is above 25 percent. Once that probability drops below 25 percent, you open the seat to discount buyers.
This single ratio drives billions in revenue decisions. Modern systems extend it to many fare classes and continuous demand, but the intuition holds: protect scarce capacity for high-value demand, but not past the point where the discount is a better bet.
Back to our oversold Friday flight. Overbooking manages no-shows, passengers who book but never arrive.
If history shows 5 percent no-shows on this route, selling 5 percent extra seats fills the plane on average. The math balances two costs:
You overbook up to the point where the marginal cost of a likely bump equals the marginal revenue of a likely extra sale. In the United States, denied boarding compensation is regulated by the US Department of Transportation, so airlines model those payout rules directly into the tradeoff.
Hotels do the same thing. A property expecting cancellations will confirm more reservations than it has rooms, then "walk" overflow guests to a comparable hotel at its own expense if everyone shows.
🎬 [VIDEO: "Revenue Management and Dynamic PricingDynamic PricingAutomatically adjusting prices in real time based on demand, competition or user behaviour to optimise revenue, margin or conversion.View full definition →" — youtube.com — a clear MIT-style walkthrough of yield management and the seat protection logic]
Fare fences and Littlewood's rule are the classic framework. Modern dynamic pricingdynamic pricingAutomatically adjusting prices in real time based on demand, competition or user behaviour to optimise revenue, margin or conversion.View full definition → updates prices continuously based on live signals:
If bookings for a specific hotel night are running ahead of forecast, the system raises rates. If a flight is booking slow with two weeks left, it releases cheaper fare classes to stimulate demand.
Hotels layer this with length-of-stay controls. A hotel expecting a sold-out Saturday may refuse a one-night Saturday booking while accepting a guest staying Thursday through Sunday, because the multi-night guest is worth more across the full period.
A word of caution for 2026: dynamic pricingdynamic pricingAutomatically adjusting prices in real time based on demand, competition or user behaviour to optimise revenue, margin or conversion.View full definition → invites regulatory and reputational scrutiny. Prices that appear to use sensitive personal data or that spike opportunistically during emergencies draw legal challenges and public backlash. Transparency and fairness are now business constraints, not afterthoughts.
Knowledge check
1. What defines inventory as 'perishable' in the context of revenue management?
2. Why is maximizing total revenue per departure a better goal than maximizing load factor?
3. Why do airlines deliberately sell more tickets than there are physical seats?
4. Select ALL correct answers about the tradeoff created by perishability in pricing decisions.
Select all the correct answers.
5. Select ALL correct answers about patterns that travel demand forecasting typically relies on.
Select all the correct answers.
Consider a 200-room city hotel facing a Saturday with a major conference in town.
1. Forecast: the model, corrected for censored demand from prior sold-out conferences, projects demand well above 200 rooms.
2. Fences: the hotel closes its cheapest advance-purchase and non-refundable rates. Only higher flexible and premium rates remain.
3. Length of stay: it requires a two-night minimum to capture the full weekend rather than filling with one-night bookings.
4. Overbooking: expecting a small cancellation rate, it confirms a few reservations beyond 200, with a walk plan ready.
5. Dynamic adjustment: as the date fills faster than forecast, rates climb further; if pace stalls, a restriction is relaxed.
Every lever serves one aim: convert a perishable, fixed set of rooms into the maximum revenue before midnight Saturday erases any unsold value.