An unsold room-night at a hotel can never be recovered. That's why hotels deliberately oversell — and treat the cancellation policy not as a cost, but as a lever.
A hotel room is what economists call a perishable product: tonight's empty room can't be sold twice tomorrow; that revenue is gone for good. At the same time, a predictable share of the reservations you confirm will cancel, or the guest simply won't arrive (a no-show). Put those two facts together and you get a classic revenue-management dilemma: if you sell only up to physical capacity, the gaps left by cancellations and no-shows will leave rooms empty even on a sold-out night.
This article covers three interconnected topics: the real scale of cancellations, why hotels deliberately oversell (overbooking), and how the cancellation policy can be turned into a demand lever.
1. The scale of the problem: the truth between "40%" and "20%"
There is no single number for the cancellation rate — and that's not a flaw, it's information.
The most-cited figure, based on D-EDGE's channel-manager data, is a ~40% average cancellation rate; the same analysis shows rates ranging 18% to 42% by channel. The channel breakdown is the strongest data in the entire debate (D-EDGE, 2024):
- GDS: ~4.6%
- Hotel website (direct): ~11%
- Booking (global group): ~37%
- Booking Holdings (Europe): ~42%
- Expedia (Europe): ~31%
The fact that direct bookings (~11%) cancel far less than OTA bookings (~37–42%) is the backbone of the "policy by channel" argument.
Against this, SiteMinder — with data from more than 44,500 hotels — reports that the average cancellation rate has fallen below 20% (~19% in 2025). So which is right? Both. The difference is largely a matter of denominator and channel mix: an OTA-heavy base pushes toward ~40%, a direct-heavy blended base toward ~20%. This apparent contradiction actually proves the most important lesson: don't trust averages until you've segmented your own data by channel, property type, and season.
The academic anchor exists too: the public dataset of 119,390 reservations published by Antonio, de Almeida and Nunes (2019) measured — even within the same chain — a 41.7% cancellation rate for a city hotel and 27.8% for a resort. Property type alone creates a 14-point gap. No-shows, meanwhile, typically run in the 1–5% band and are more predictable than cancellations.
2. Why hotels deliberately oversell: the math of overbooking
Overbooking is not a mistake, it's an optimization decision. It means selling in advance the "phantom" gaps that cancellations and no-shows will create.
The decision rests on balancing two asymmetric costs — the classic newsvendor model:
- Overage cost (Co): You oversold and everyone showed up — you have to walk a guest (relocation to another hotel, transport, comp, reputation, and loss of customer lifetime value).
- Underage cost (Cu): You didn't oversell enough — the room stayed empty, i.e., its opportunity cost / lost margin.
The optimal overbooking level is set by the critical ratio: F(Q) = Cu / (Cu + Co). Because walking a guest (Co) usually costs far more than an empty commodity room (Cu), this ratio is low — meaning hotels should oversell conservatively*. In practice the industry typically runs 2–10% above capacity.
The logic originated with airlines. Littlewood (1972) produced the first newsvendor-style booking rule for two fare classes at BOAC. American Airlines operationalized it after the 1978 deregulation; Smith, Leimkuhler and Darrow's (1992) "Yield Management at American Airlines" in Interfaces won the 1991 Franz Edelman Award. Sheryl Kimes carried the idea into hospitality (Cornell, 1989 and 2003), laying the foundation for "yield management" and "revenue management" in the lodging sector.
3. The real cost of "walking" a guest
The dangerous side of overbooking lies precisely in this asymmetry. An empty commodity room costs one night's margin; but a walked guest can cost far more.
The walk cost is not just that night's room: it's the alternative hotel's rate, transport, comp, the risk of a negative review, and — most importantly — the loss of all of a loyal guest's future bookings. In the US there is no federal law requiring walk compensation; practice is brand-driven. For top-tier loyalty members, compensation can be substantial (for example, hundreds of dollars plus tens of thousands of points per incident at some luxury brands). This is exactly why the optimal overbooking level is kept low: the Co ≫ Cu asymmetry financially punishes aggressive overbooking.
4. The cancellation policy is a lever
A cancellation isn't just something that happens to you — it's largely something you design. Policy is the most powerful lever that moves both volume and cancellations at once.
Flexible vs non-refundable. Fully flexible rates cancel at over 35%, while non-refundable rates cut that sharply — but attract fewer guests. The cleanest evidence of the two-sided tradeoff is Booking.com's own data: adding a non-refundable rate plan reduces cancellations by at least 9% and increases bookings by at least 5%; but offering only non-refundable rates lowers total bookings. The right answer is not a single policy but a rate portfolio.
Price itself triggers cancellations. The strongest academic evidence is Urrea, Huang and Zhang (2026, Cornell Hospitality Quarterly): at a high-end US hotel, 20.8% of 15,101 reservations were canceled, and every $50 increase in booking price raised the cancellation hazard by 16%. About 27% of the total effect is explained by guests who keep monitoring prices after booking and cancel when the price drops. Ignoring this link can cost up to 11% of high-season revenue.
"Book now, cancel later, rebook cheaper." A 2026 study examining ~2.2 million reservations (International Journal of Hospitality Management) found that every $10 drop in the rebooking price increases cancellation probability by 4.7%; this behavior caused €1.24M in gross revenue displacement in the sample. The suggested guardrail: cap downward price moves within 14 days of arrival.
Deposits and prepayment. Prepaid/deposit bookings cancel roughly 50% less than average. Tightening the free-cancellation window to 24–48 hours before check-in raises commitment without removing flexibility entirely.
5. No-show handling and practical conclusion
No-shows are more predictable than cancellations: they arrive as a stable base rate per segment and season (~1–5%) and feed directly into your overbooking limit. The main levers are credit-card guarantees, deposits, and prepayment — they shift the cost of not showing onto the guest.
So what should a hotelier do?
- Forecast net, not gross, demand: Expected net arrivals = gross confirmed bookings − predicted cancellations − predicted no-shows. Compute this per segment/channel/season, not on a blended average. The strongest predictors: lead time, deposit type, and channel (Antonio et al.).
- Derive the overbooking limit from the cost asymmetry: oversell up to the critical ratio
Cu/(Cu+Co); because the walk cost (Co) dwarfs an empty commodity room (Cu), keep limits conservative. - Tier policies by channel and lead time: flexible (with a short 24–48h free window) to capture demand; non-refundable (at ~8–12% discount) to lock committed guests; deposits for high-cancellation segments.
- Protect reputation: never oversell into dates with loyalty or special-occasion concentration. The gain from a filled room never covers the cost of lost trust.
Summary: Cancellations, no-shows and overbooking are three faces of the same coin: all are about managing the gap between "confirmed reservation" and "revenue that will actually materialize" when you work with perishable inventory. The right approach is not a single rigid rule but a disciplined system that forecasts net demand instead of gross, derives overbooking from the cost asymmetry, and designs the cancellation policy as a portfolio. An empty room doesn't come back — but neither may a wrongly-walked guest.
Sources
- Antonio, N., de Almeida, A. & Nunes, L. (2019). Hotel booking demand datasets. Data in Brief, 22, 41–49. (119,390 reservations; city hotel 41.7%, resort 27.8% cancellation)
- Antonio, N., de Almeida, A. & Nunes, L. (2019). Big Data in Hotel Revenue Management: Exploring Cancellation Drivers. Cornell Hospitality Quarterly.
- Urrea, G., Huang, X. & Zhang, D. (2026). The Effect of Pricing on Cancelations at a High-End U.S. Hotel. Cornell Hospitality Quarterly. ($50 → 16% higher cancellation hazard; 20.8% canceled; up to 11% of high-season revenue lost)
- Cancel, rebook, save: Revenue leakage from price cuts in hotels. (2026). International Journal of Hospitality Management. (2.2M reservations; $10 → 4.7%; €1.24M displacement)
- Smith, B., Leimkuhler, J. & Darrow, R. (1992). Yield Management at American Airlines. Interfaces, 22(1):8–31. (1991 Franz Edelman Award)
- Littlewood, K. (1972). Forecasting and control of passenger bookings. (First newsvendor booking rule)
- Kimes, S. E. (1989). The Basics of Yield Management and (2003) Revenue Management: A Retrospective. Cornell Hotel & Restaurant Administration Quarterly.
- D-EDGE — Hotel Distribution Report 2024 and no-show/cancellation prevention guide (channel-level cancellation table; no-show 1–5%; prepaid ~50% fewer cancellations)
- SiteMinder — Hotel Booking Trends (blended cancellations <20%; global ~19%)
- Booking.com for Partners — Managing cancellations (non-refundable plan: cancellations −9% / bookings +5%)