The no-show revolving door

Every booking-based business — restaurants, hotels, clinics, salons, professional practices, rentals — knows the phenomenon: a share of customers book and do not show up. The instinctive reaction is moral («rude people») or defensive (deposits, reminders). Both legitimate, but neither answers the economic question: how many bookings should I accept, knowing that a share will evaporate?

The question has a computable answer, and it is one of the most elegant in operations: the same mathematics airlines have used for half a century, reduced to the four numbers every operator knows — capacity, no-show rate, revenue of an occupied seat and cost of a turned-away customer.

This calculator applies it to your dining room, your hotel or your schedule: it says what your current booking policy costs you today, where the break-even sits, and how much «denial» risk is rational to accept. Everything per service, in your currency, with a dimensionless verdict.

The empty seat is a certain cost, the denial a possible one

The model's heart is an asymmetry instinct rejects and arithmetic imposes. If you accept only as many bookings as you have seats, the no-shows certainly leave you empty seats: with 50 covers and a 12% absence rate, six tables per service, night after night. If you accept a few extra bookings, you risk — it is not certain — having to turn someone away when everyone shows up.

Certain cost against possible cost: it is the structure of the newsvendor problem, and the solution has a name — the critical ratio CR = R/(R+D). With 80 of revenue per seat and 150 of cost per denial, it is 34.8%: meaning it is economically rational to push overbooking up to the point where the probability of turning someone away reaches 35%. Not 0%.

The sentence deserves rereading, because it reverses common sense: zero overbooking is not prudence, it is a choice with a cost — the certain cost of empty seats — which must be weighed against the risk one wants to avoid. The calculator makes exactly that comparison.

The same dining room, five policies

The table applies the same room — 50 covers, 12% no-show, 80 revenue, 150 denial cost — to five booking policies. The room is the same in every row: only how many bookings you accept changes.

The cost column draws a U: high on the left from empty seats, high on the right from denials, with the minimum near the break-even bookings B* = C/(1−s) = 56.8. The «prudent» policy of the first row — by far the most widespread — is the second most expensive in the table: 480 per service, which over 300 services a year makes 144,000 of capacity paid for and not sold.

And the minimum is not an arbitrary point: it is where expected shows exactly equal capacity. Below it you are buying empty seats; above it, denials. The U says one more thing: the errors are not symmetric — with these numbers, overshooting by three bookings costs less than undershooting by three, because the critical ratio is high.

How this calculator works

The capacity per service is the unit you sell: covers per seating, rooms per night, chairs per time band, schedule slots. The no-show rate is the share of bookings that neither shows nor cancels in time: your booking system knows it, and it is worth measuring by day and by channel, because it is far from uniform.

The revenue per occupied seat is the margin a filled seat generates — the average ticket net of variable costs, not the list price. The cost per denied customer is the hardest and most important item: what you spend to remedy (voucher, upgrade, comped dinner, rebooking elsewhere) plus an honest estimate of the relationship damage. It must be greater than zero: if turning someone away cost nothing, there would be no rational limit to overbooking.

The bookings you accept today are your current policy — the input is your choice, and the main result evaluates it: you can move it and see the table's U with your own numbers.

How to read the results

The average cost of your policy is the expected sum of the two wastes — empty seats at their revenue, denials at their cost — per service. The break-even bookings B* are the level at which expected shows equal capacity: the U's minimum in the deterministic model.

The critical ratio is the risk compass: the denial probability up to which it is rational to push, given your R and D. The overbooking index ρ = B/B* is the verdict: below 0.75 you are in the revolving door — giving capacity away; between 0.75 and 1 the leak shrinks but remains; between 1 and 1.15 you are in the break-even zone, where variance decides; beyond 1.15 you are systematically denying, and denial is a fixed cost in disguise.

Why break-even is not the destination

This page's model is deterministic: it reasons on averages. But actual attendance oscillates around the mean, and the last word belongs to variance — which is why the break-even zone's verdict names it explicitly.

The practical rule for handling it is the critical ratio: if CR is high (denial cheap relative to the seat), you can sit at break-even or slightly above, because occasional overruns cost little; if CR is low (denial very costly), it pays to stop a few bookings below B*, sacrificing some leak in exchange for short queues on bad days.

There is also a path the calculation suggests without saying it: reduce s instead of raising B. Every point of no-show removed moves the break-even toward capacity and shrinks the problem at the root — and it is the only lever that improves both sides of the U at once. We return to it among the strategies.

The newsvendor family

This calculator belongs to a family of problems identical under different clothes: how much stock to hold when demand is uncertain and erring up costs differently from erring down. The safety stock calculator is the same problem seen from the warehouse: there the «empty seat» is the stock-out and the «denial» is the capital tied up.

The kinship is not a curiosity: it is a method. Whoever has understood the critical ratio once recognizes it everywhere — in sizing a buffet, in staffing a shift, in a switchboard's phone lines, where the Erlang C calculator applies the queueing version of the same logic.

Overbooking's difference is that the uncertainty sits on the side of demand already sold: the customer has promised, and the promise only holds statistically. That is what makes the problem culturally thorny — it seems to be about trust — and mathematically trivial: a promise that shows up 88% of the time is worth 0.88 seats, and must be counted as such.

Where the no-show rate comes from

The whole model rests on s, and the good news is that it is the best-measured parameter in the whole operation: every booking system records it. The bad news is that almost nobody looks at it disaggregated — and the average deceives.

The dimensions that matter: the channel (anonymous portal bookings have multiples of the no-show rate of direct regulars), the day and hour (Saturday dinner and Monday lunch are different populations), the lead time (the booking made three weeks ahead evaporates more than the same-day one) and the party size. A serious overbooking policy uses an s per segment, not one per venue.

And the rate is not a natural constant: it is a managed variable. Automatic reminders, card guarantees, deposits on groups, an active waitlist — each lever moves s by points, and every point moves the break-even. The calculator lets you test the scenarios in ten seconds.

Limits of the estimate

The model is deterministic on averages: it does not compute your policy's exact denial probability, which depends on the attendance distribution (approximately binomial) and on the correlation between bookings — groups show or fail together, which widens the variance versus the independent case.

It also assumes a no-show rate independent of the overbooking level and customers homogeneous in revenue: in reality it pays to overbook more in low-value segments and protect the best tables. It does not model dynamic effects — a customer denied today books less tomorrow, and a waitlisted customer recovered is a denial avoided for free.

A parametric estimate for information purposes: it supports the calibration of your booking policy, it replaces neither the data in your booking system nor, in regulated sectors such as air transport, the statutory compensation obligations.

The no-show ladder

The first table holds the «prudent» policy fixed — 50 seats, 50 bookings accepted, 80 of revenue — and raises the no-show rate. It is a photograph of what doing nothing costs at different absenteeism levels.

The progression is linear but the reading is not trivial: at 5% the revolving door costs 200 per service and one can reasonably decide it is not worth the complexity of an overbooking policy; at 20% it costs 800 — four times as much — and ignoring it means leaving the equivalent of ten full covers on the table. The no-show rate alone decides whether this problem deserves your attention: below 5% it is a nuisance, above 15% it is a P&L line.

The last column tells the other half: the higher the rate, the further break-even moves from capacity. At 30% you would have to accept 43% more bookings than seats — levels no dining room manages serenely. It is the signal that beyond a certain threshold the right answer is no longer «overbook more» but «bring the rate down»: deposits, confirmations, cancellation policies.

The denial-cost ladder

The second table holds revenue fixed — 80 per seat — and raises the cost of the denied customer: from fast casual where the turned-away guest gets a voucher, to contexts where a denial is a serious wound to the relationship.

Revenue is the same in every row: what moves the critical ratio is only how much saying no costs. And here the model stops being mathematics and becomes strategy: the denial cost is not a fact of nature, it is partly a choice. Whoever builds a good «plan B» for the turned-away customer — the reserved table at the sister venue, the upgrade, the comped dinner with a fast lane at the next visit — is literally buying the right to overbook more.

The 600 row deserves a note: critical ratios below 15% describe contexts — medical practices with fragile patients, services with high reputational value — where aggressive overbooking is economically wrong as well as commercially risky. The model does not always push toward audacity: with these numbers, it prescribes caution.

The airline twin: why carriers pull it off

Airline overbooking is the most studied example because there every quantity in the model is contractualized. The «cost per denied customer» is not an estimate: it is written in the regulations — the monetary compensation of European regulation 261/2004, the US DOT rules on denied boarding, the Canadian APPR protections — and the carrier knows it to the cent before deciding how many tickets to sell.

There is a second, less visible lesson: airlines turned the denial from a suffered event into a volunteer auction. Asking who is willing to leave later in exchange for a voucher means buying the seat back from the passenger who values it least — and the denial cost collapses, because it is no longer the average passenger's damage but the most flexible passenger's price. It is the industrial version of the previous section's «plan B», and it works at any scale: even a well-run waitlist is a silent auction.

The third lesson is organizational: airlines do not use one no-show rate, they use thousands — by route, day, hour, fare class, purchase lead time. No dining room needs that granularity, but the direction is the same: two or three well-chosen segments beat any overall average.

Strategies to lower the no-show rate

The calculator shows what each point of rate removed is worth: the no-show table above says moving from 12% to 10% is worth 80 per service — 24,000 a year over 300 services. With that number in hand, the levers can finally be priced instead of debated.

The automatic reminder (SMS or message the day before, asking for confirmation) is the lever with the best documented cost-effect ratio: it costs cents and in typical settings removes whole points. The card guarantee with no upfront charge raises the psychological threshold of absence without real economic friction. The deposit is the most powerful and most costly lever: it nearly zeroes no-shows but repels a share of customers at booking — reserve it for the segments where the damage is greatest, large parties and peak nights.

The easy cancellation policy is the counterintuitive lever: making it trivially simple to cancel turns silent no-shows into timely cancellations, and a cancellation with notice is a resellable seat, that is, a non-problem. Whoever makes cancelling hard to «protect themselves» is manufacturing their own no-shows.

Segment overbooking: whom to overbook

Applying a single coefficient to all demand is the crude version of the model. The fine version exploits the fact that different segments have different rates: the table booked through an anonymous portal with a 25% rate and the regular who calls directly with a 3% rate do not deserve the same treatment.

The operating rule is to overbook where the rate is high and the denial damage is low: anonymous last-minute bookings, segments with no history. And to protect — that is, not overbook — where the rate is low and the damage is high: regular customers, special-occasion tables, bookings with special requests. The apparent paradox is that the best customer gets the most prudent policy: it is exactly the right order of things.

In practice two classes suffice: a share of «guaranteed» seats sized on reliable demand, and a «flexible» share where the overbooking coefficient works on anonymous traffic. The calculator is then used twice, once per class, with the respective rates — and the sum of the two policies systematically beats the single average.

Mistakes that inflate the bill

The first is using the annual average rate: the average between the rainy Tuesday and the peak Saturday describes neither. Overbooking is calibrated per service, not per venue — and high-demand services, where the empty seat costs most, often have lower rates because the clientele is more motivated.

The second is counting gross revenue instead of margin: the empty seat does not lose you the average ticket, it loses you the ticket minus the variable cost of serving it. Inflating R inflates the critical ratio and pushes you to overbook more than you should.

The third is underestimating the denial cost by stopping at the immediate remedy: the 50 voucher is not the cost if the turned-away customer never returns and tells the story. The honest estimate includes the relationship — and it is almost always higher than what you spend that same evening.

The fourth is treating break-even as a fixed target: B* moves with the rate, and the rate moves with seasons, channels and your own countermeasures. The policy must be recalibrated when the ingredients change — ten seconds with this calculator — not set once and forgotten.

Frequently asked questions

What are the break-even bookings?

The level B* = capacity / (1 − no-show) at which expected shows exactly equal the seats: with 50 seats and a 12% no-show it is 56.8. Below it you buy empty seats, above it denials.

What is the critical ratio?

CR = R/(R+D): the seat's revenue divided by the sum of revenue and denial cost. It indicates the denial probability up to which pushing is rational: with 80 and 150 it is 34.8%.

What is the overbooking index?

ρ = B/B*: the bookings you accept divided by the break-even ones. Below 0.75 you give capacity away; between 1 and 1.15 you are in the break-even zone; beyond 1.15 you deny systematically.

Why does accepting only as many bookings as seats cost money?

Because no-shows leave seats empty with certainty: at 12% on 50 seats that is 6 per service, i.e. 480 of lost revenue — 144,000 a year over 300 services. A certain cost traded against a risk that can only partly be avoided.

What goes into the cost per denied customer?

Everything you spend to remedy — voucher, upgrade, comped dinner, rebooking elsewhere — plus an honest estimate of the relationship and reputation damage. It is the hardest input, and it determines how much risk is rational.

How do I measure the no-show rate?

From the booking system: bookings neither shown nor cancelled in time, divided by the total. Measure it by channel, day, lead time and party size: the venue-wide average deceives, and a serious policy uses a rate per segment.

Does the model work for hotels, clinics and salons too?

Yes: only the capacity unit changes — rooms, schedule slots, chairs. The break-even and critical-ratio mathematics is identical, and it is the same airlines have used for half a century.

What if I never want to turn anyone away?

A legitimate choice with a visible price: the certain cost of empty seats. The calculator quantifies it — and shows it can be reduced even without overbooking, by lowering the no-show rate with reminders, guarantees and waitlists.

Why is the deterministic break-even not enough?

Because attendance oscillates around the mean: at break-even, you overflow half the days. If denial is cheap (high CR) that is fine; if it is costly (low CR) it pays to stop a few bookings below B*.

Does a waitlist replace overbooking?

It completes it: every customer recovered from the list is an empty seat filled with no denial risk. The most robust combination is moderate overbooking plus an active waitlist for the bad days.

Do deposits not solve everything?

They reduce s — often a lot — but they have a commercial cost: booking friction and customers lost to no-deposit competitors. They must be priced like any lever: the calculator shows what every point of no-show removed is worth.

Does this tool replace revenue management?

No. It gives the decision's economic structure for information purposes: a revenue management system adds distributions, seasonality and dynamic pricing. But the compass — break-even and critical ratio — is the same.

Updated July 2026