The house count says sold out plus six. It is four in the afternoon, the arrivals are trickling in, and somewhere in the back of my head a small calculator is running: how many of the people who booked tonight will actually show. Because if the usual handful of no-shows materializes, sold out plus six lands exactly full. If everyone shows, I am walking six guests I never wanted to walk.

That tension is overbooking, and it is not recklessness. It is a calculated bet that a predictable number of booked rooms will not turn into occupied rooms. The question is never whether to oversell. Any hotel that refuses to oversell leaves money on the floor every single night. The question is how much, and that number comes from math you can actually do.

Why you oversell at all

A hotel room is perishable in the purest sense. Tonight's empty room cannot be sold tomorrow. When the clock runs out, an unsold room earned nothing and never will, and the cost of building and staffing it was already spent. That is the same brutal logic I laid out in why a dollar of rate beats a point of occupancy: once the room exists and the desk is staffed, an empty room is pure loss.

Now layer on reality. Some of the guests who booked tonight will not come. People no-show. People cancel late. People check out a day early. If you only ever sell to your exact room count, every one of those gaps becomes an empty room you could have sold, on a night you were technically sold out. You turned away paying guests to protect rooms that ended up vacant anyway.

Overbooking closes that gap. You sell past your physical count by roughly the number of rooms you expect to come open through no-shows and early departures, so that when the attrition happens, you land full instead of short. Done right, it is the difference between a sold-out night that finished at 96 percent and one that finished at 100.

The two costs you are balancing

Every oversell decision is a bet between two costs, and naming them plainly is most of the work.

The first is the cost of an empty room. If you undersell and a room sits vacant, you lose that night's revenue, minus the small variable cost you did not incur by not having a guest. Call it close to the room's rate. On a strong night at a high rate, an empty room is an expensive miss. On a soft night, less so.

The second is the cost of a walk. If you oversell and everyone shows, you are out of rooms and you have to send a guest to another hotel. That cost is real and it is bigger than one night's rate. You pay for their room at the other property, often at a premium because you are buying last minute. You usually cover transportation. Sometimes you comp a future stay to make it right. And there is a cost that never shows up on the folio: the guest who was walked may never come back, and may tell everyone they know. Walking a guest well takes real care, which is a craft of its own, one I broke down in how to walk a guest with grace.

The whole game is that these two costs are not equal. A walk usually costs several times what an empty room costs. That asymmetry is the single most important fact in overbooking, and it is why you never oversell all the way up to your expected no-shows.

The formula, in plain terms

Here is the core idea without the intimidating notation. You keep overselling one more room as long as the expected cost of that extra sold room is lower than the expected cost of leaving the equivalent room empty.

Each additional room you oversell has a chance of causing a walk, equal to the probability that fewer no-shows happen than you bet on. Each room you decline to oversell has a chance of ending the night empty, equal to the probability that the no-shows do happen. You push the oversell number up until the expected cost of one more walk risk equals the expected cost of one more empty-room risk. That balance point is your optimal oversell level.

Because a walk costs so much more than an empty room, the balance tips well before you reach your full expected no-show count. A quick way to feel it: if a walk costs roughly three times what an empty room costs, you only oversell to the point where the chance of walking is about a third of the chance of an empty room. So if you expect eight no-shows on a given night, you might oversell by five, not eight, keeping a buffer against the nights when the no-shows do not show up to be absent.

Let me put illustrative numbers on that, and these are illustrative, not any property I have run. Say an empty room costs you $200 in lost rate and a walk costs you $600 all in. The walk is three times as expensive, so you protect against it three times as hard. You oversell to where the odds of actually walking someone stay low, and you stop well short of the maximum, because the downside is lopsided against you.

Where the no-show number comes from

The whole formula rests on one estimate: how many booked rooms will not become occupied rooms tonight. Guess that badly and the rest is noise. Estimate it well and overbooking becomes routine rather than nerve-wracking.

You build the no-show rate from history, and you build it by segment, because segments behave nothing alike.

  • Non-refundable and advance purchase bookings barely no-show. The guest already paid, so they show or eat the cost. Treat these as near-certain arrivals.
  • Guaranteed transient on a credit card no-shows at a modest, fairly stable rate you can pull from your own data.
  • Non-guaranteed and hold-until reservations no-show heavily. These are the rooms most likely to evaporate.
  • Corporate and negotiated rates vary by account, and some are reliable enough to plan around.
  • Groups have their own attrition pattern, and blocks that never fully pick up are their own source of open rooms.

Day of week matters, lead time matters, and the season matters. A Tuesday business night and a Saturday leisure night have different no-show behavior, and a room booked three months out is a different risk than one booked yesterday. The point is that "expected no-shows" is not one number for the hotel. It is a blend of segment rates applied to tonight's actual booking mix, which is why the same 100 percent-plus house count can be a safe oversell one night and a reckless one the next.

The buffer and why you never oversell to the max

Even with a good no-show estimate, you hold a buffer. You do not oversell right up to your expected no-shows, because the estimate is an average and any given night can land on the wrong side of it. The night everyone decides to show up is the night you did not want to be at your theoretical limit.

The buffer is just the asymmetry made concrete. Since a walk hurts so much more than an empty room, you accept a slightly higher chance of an empty room in exchange for a much lower chance of a walk. That is not timidity. That is pricing the downside honestly. An operator who oversells to the absolute edge to hit 100 percent every night will eventually have a night that shows in full, and the cost of that one blowout, in walks and in goodwill, can swallow the small gains from all the nights they squeezed.

This is also where the number stops being purely mathematical and starts being judgmental. The formula gives you a range. Where you land in that range depends on how much risk the night can absorb, and that leads straight to the live picture.

Reading the count as the night comes in

The oversell number you set days out is a starting position, not a verdict. It moves as the night arrives, and moving it well is a matter of reading the house count as the picture sharpens.

Through the afternoon you are watching the arrival pace against what you expected. If arrivals are coming in hot and the no-shows are not materializing, you tighten, stop taking walkins, and start working the phones for outbound relocations before you are forced into them at midnight. If the expected no-shows are showing up as no-shows on schedule, you relax, and you can even take a late walkin at a premium into a room you now know is coming open.

This live adjustment is the difference between managing a full house and being managed by it. The math sets the bet. The house count tells you, hour by hour, whether the bet is landing, and handling a sold-out night is really the discipline of adjusting that bet in real time instead of praying on it.

When the math says walk, walk on purpose

Sometimes, despite good math and good watching, the no-shows do not come and you are genuinely out of rooms. When that happens, the worst move is to freeze and let it become a fight at the desk at eleven at night. If a walk is coming, you decide early, you choose who is walked by who is easiest to relocate and least disrupted, and you make it right generously.

A walk handled early and well, with a booked room at a comparable hotel, transport arranged, and a sincere gesture toward the guest's next stay, is a recoverable moment. A walk handled late and defensively is a lost guest and a story that travels. The overbooking math is what makes walks rare. The care in the moment is what keeps a rare walk from costing you the relationship. Both matter, and the fuller version of the operational side lives in the overbooking playbook.

What I want a team to understand

I do not need a front desk agent to run probabilities in their head. I need them to understand three things. First, we oversell on purpose, because empty rooms on a sold-out night are a real loss, not a comfort. Second, we oversell carefully, with a buffer, because a walk costs far more than an empty room. Third, the number is not fixed; it breathes with the arrival pace, and the desk's read of the room through the evening is part of the calculation, not separate from it.

When a team gets this, overbooking stops feeling like gambling and starts feeling like what it is: a routine bet with a known edge and a managed downside. The goal is not the thrill of a sold-out house. It is the quiet outcome of a hotel that finishes full most nights and almost never walks a soul.

The takeaway

Overbooking is not nerve, it is arithmetic. Estimate honestly how many booked rooms will not become occupied rooms, oversell by less than that so the walk you dread stays unlikely, and let the number move with the arrival pace instead of freezing it days out. Get the math right and the empty rooms disappear. Keep the buffer honest and the walks stay rare. That is the whole trade: a full house most nights, and a walked guest almost never.

Questions from the desk

How do hotels calculate how much to overbook?

They estimate expected no-shows, cancellations, and early departures from historical data by segment, then oversell by less than that number, holding a buffer. The buffer size is driven by how much more a walk costs than an empty room. You push the oversell up until the expected cost of a walk risk equals the expected cost of an empty-room risk.

Why do hotels overbook at all?

Because rooms are perishable and a predictable share of booked guests never arrive. Selling only to the exact room count means every no-show becomes an empty room you could have sold on a night you were sold out. Overbooking recovers that lost revenue.

How do hotels avoid walking guests when they overbook?

By estimating no-shows conservatively, keeping a buffer below the expected number, and watching arrival pace through the evening so they can tighten early if the no-shows are not showing up. When a walk is unavoidable, deciding early and relocating generously keeps it rare and recoverable.

Does overbooking cost more than it earns?

Only when it is done to the edge. Because a walk costs several times what an empty room costs, overselling to the theoretical maximum eventually produces a blowout night that erases the gains. Done with a buffer, overbooking earns steady, low-risk revenue that outweighs the occasional walk.