A booking deposit is a trade, not a repair: you give up some guests who refuse to book at all in exchange for fewer guests who book and never arrive. It pays while refusals stay under a threshold set by your own contribution per booking, the drop in no-shows you actually measure, and the amount you keep.
What a deposit actually trades: no-shows for refusals
Every argument about deposits collapses into one sentence: it removes some no-shows and creates some refusals, and nobody knows which side is bigger until both are counted.
That is the structure of the arithmetic, not a figure of speech. Without a deposit, a booking request becomes either a booking that arrives or one that does not. With a deposit a third branch appears before both: the guest hears the condition and hangs up. That is not a no-show — it is a booking that never existed, and no report will ever count it.
One side of the trade is measured for you; the other is invisible. Anyone arguing from the visible side alone concludes that a deposit is free money.
Refusal rate — the share of would-be bookings that do not happen because a deposit is required. It is measured on booking attempts, not on bookings taken: the denominator has to include the people who walked away.
What this page does not do
It does not measure no-shows. The measure and its two bases belong to the table turnover page, and the confirmation chain that prevents them to the no-show article. Here the no-show rate is an input you already have, used to answer one decision.
It makes no legal claim about whether you may keep the money either — that depends on your terms, and a section below shows what the Polish Civil Code does and does not say.
And it names no industry norm, because there is none. The Eurostat dissemination catalogue, read in full on 27.08.2026, holds no dataset on booking absence: a sweep of its 1 971 713 bytes returns zero occurrences of "no-show" and of "cancellation", three of "reservation" — all the same table about online appointments — and one of "deposit", a bank interest rate. The Polish statistical office is no closer: its trade section lists twenty-six items, one a methodological guide to trade and catering activity, and none mentions reservations, absences or deposits.
The input you already have, quoted from where it is measured
The no-show rate is not defined here. It is defined and measured on the turnover page, and here it is, quoted:
No-show rate — the second measure of the same absence, counted not in seat-hours but in bookings: No-show rate % = Missed bookings ÷ Total bookings taken × 100, where missed bookings are those where nobody arrived and nobody cancelled, and total bookings taken is every booking accepted for the same service, kept and missed together. Both counts are in bookings, the result is a share in per cent. Quoted from the table turnover page.
Two details in it decide whether your own number means anything.
That is not rounding: dividing by the bookings that worked flatters the number exactly when it matters most.
And the rate is not the loss. The same page keeps a second measure — no-show seat hours — because a rate cannot say what an absence costs.
What a seat-hour is worth in money is a separate matter, held by the RevPASH page.
You need both, and in different places. The deposit changes how often bookings fail, so the rate is the variable that moves; what a failure costs is money per booking, and that is the other input below. Substitute one for the other and the arithmetic runs perfectly on the wrong question.
The contribution one booking attempt is worth without a deposit
The second input is what a booking is worth when it works: not the bill, but the part that stays with you.
Contribution margin — quoted from the break-even page, which owns the term: "the money left from net sales after variable costs, available to cover fixed costs and, once they are covered, to become profit." Per booking that is the whole party at one table for one visit — the c used here. What such a booking costs to acquire belongs to the cost per booking page.
Expected contribution per booking attempt — the contribution of one booking weighted by the probability that it happens at all and the probability that the guest arrives. It is the only unit in which a deposit and no deposit compare: a deposit changes the number of bookings, and per-booking figures hide exactly that.
Expected contribution per attempt without a deposit = (1 − No-show rate) × Contribution margin per booking
No-show rate— your measured share of bookings where nobody arrives, without a deposit, a dimensionless share between 0 and 1;Contribution margin per booking— the contribution a booking leaves when the party arrives, PLN per booking;- the result is PLN per booking attempt: a dimensionless share multiplied by PLN per booking.
Work it through with numbers that belong to nobody but this example.
c = 120 PLN per booking and the 7.0 % no-show rate from the Friday above, so p = 0.07.Then (1 − 0.07) × 120 = 111.60 PLN per attempt. Backwards: 111.60 ÷ 120 = 0.93, and 1 − 0.93 = 0.07.
A booking request is worth less than a booking, and the gap between 120 and 111.60 is what the room already pays for absence before anyone proposes anything.
The same arithmetic with a deposit: three inputs change at once
A deposit does three things at once, and every failed argument about deposits noticed one of them and not the others.
- Fewer bookings happen. Some guests refuse the condition. That is the refusal rate,
r. - Of the bookings that do happen, fewer fail. That is a new no-show rate,
p′, and it is a measurement, not a hope. - When one does fail, you keep something. That is the recovered amount,
D.
Recovered amount — what the venue keeps when a booking with a deposit does not arrive. It is not the contribution that was lost and is normally smaller: a deposit sized to be acceptable to a guest rarely covers the evening the empty table would have produced.
Expected contribution per attempt with a deposit = (1 − Refusal rate) × [ (1 − No-show rate with deposit) × Contribution margin per booking + No-show rate with deposit × Recovered amount ]
Refusal rate— the share of booking attempts that end without a booking because a deposit was asked for, a dimensionless share;No-show rate with deposit— your measured share of absences among bookings that were made with a deposit, a dimensionless share;Contribution margin per booking— the samecas above, PLN per booking;Recovered amount— what you actually keep from a deposit when nobody arrives, PLN;- inside the bracket: a share times PLN per booking, plus a share times PLN, equals PLN per booking taken. Outside: multiplied by a dimensionless share, giving PLN per booking attempt — the same unit as the formula above, which is the only reason the two can be compared at all.
Continue the example. After introducing a deposit you measured p′ = 0.02, and you keep D = 40 PLN. The bracket is 0.98 × 120 + 0.02 × 40 = 117.60 + 0.80 = 118.40 PLN per booking taken — what a booking is worth now, before subtracting anyone who refused.
Everything now hangs on how many refused.
0.97 × 118.40 = 114.85 PLN per attempt, against 111.60 without the deposit.0.90 × 118.40 = 106.56 PLN, and the deposit has cost you money.Between three and ten lies a rate where the two are equal, and that rate is the answer to the whole page.
The break-even refusal rate: how many guests you can lose without losing money
Set the two expressions equal and solve for the refusal rate.
Break-even refusal rate = 1 − [ (1 − p) × c ] ÷ [ (1 − p′) × c + p′ × D ]
p— no-show rate without a deposit, a dimensionless share;p′— no-show rate with a deposit, measured on the same kind of service, a dimensionless share;c— contribution margin per booking, PLN;D— recovered amount kept when a booking with a deposit fails, PLN;- numerator and denominator are both PLN per booking, so their ratio is dimensionless; one minus a dimensionless number is a dimensionless share, and the answer is a share of booking attempts.
0.93 × 120 = 111.60; denominator 0.98 × 120 + 0.02 × 40 = 118.40; ratio 111.60 ÷ 118.40 = 0.942568; threshold 1 − 0.942568 = 0.057432, that is 5.74 %.0.942568 × 118.40 = 111.60 PLN per attempt — identical to no deposit, to the last grosz.So here the deposit pays while fewer than six attempts in a hundred walk away because of it. Above that it costs money silently, in a line item that does not exist.
The threshold does not depend on the size of the deposit — only on its ratio to the booking
Divide numerator and denominator by c and it becomes 1 − (1 − p) ÷ [ (1 − p′) + p′ × (D ÷ c) ]: the two money figures appear only as their ratio.
D ÷ c = 40 ÷ 120 = 0.3333, and 1 − 0.93 ÷ 0.986667 = 0.057432 — the same 5.74 %.The section on party size spends that fact.
Why the threshold collapses to zero when the deposit changes nothing
The formula has a degenerate case, and it belongs in the open: without it a deposit reads as unconditionally clever.
Put p′ = p and D = 0. The numerator is (1 − p) × c; the denominator is (1 − p) × c + p × 0, the same number. The ratio is exactly 1 and the threshold exactly 0 %. A deposit that does not lower no-shows and holds nothing back does not pay for one refusal — not "pays off slowly", a pure loss from the first guest who declines.
Two neighbouring cases say where the value sits.
| Case | p′ | D | Break-even refusal rate | What it means |
|---|---|---|---|---|
| The deposit changes nothing and keeps nothing | 0.07 | 0 PLN | 0 % | Every refusal is a straight loss |
| It keeps money but changes no behaviour | 0.07 | 40 PLN | 2.45 % | The kept money alone buys a little room |
| It changes behaviour but you refund everything | 0.02 | 0 PLN | 5.10 % | Almost the whole effect, without keeping a grosz |
| It does both, as in the example above | 0.02 | 40 PLN | 5.74 % | The full effect |
Read the third row against the fourth: of the 5.74 points of tolerance, 5.10 come from behaviour and 0.64 from money kept. A deposit is a commitment device with a cash component, not a cash device — and without a measured p′ nothing says you are in any row but the first.
p′ is the one measurement you cannot skip. p you already have, c you build from your own figures, D you set. But what your no-show rate becomes with a deposit in place exists only if you introduce it somewhere narrow — one service, one table size — and measure. Anyone quoting it without having run it is quoting a rumour.
What a deposit really recovers: money kept is not revenue saved
The commonest mistake here is treating the amount kept as the loss avoided.
| Step | Without a deposit | With a deposit, 3 % refuse |
|---|---|---|
| Booking attempts | 100 | 100 |
| Bookings taken | 100 | 97 |
| Bookings that fail | 7.00 | 1.94 |
| Parties that arrive | 93.00 | 95.06 |
| Contribution from arrivals | 11 160.00 PLN | 11 407.20 PLN |
| Deposits kept | — | 77.60 PLN |
| Total | 11 160.00 PLN | 11 484.80 PLN |
| Per booking attempt | 111.60 PLN | 114.85 PLN |
The other three quarters came from parties that turned up and ate.
On the failures themselves the arithmetic is exact: the 1.94 bookings that still failed cost 1.94 × 120 = 232.80 PLN of contribution against 77.60 PLN kept, and 77.60 ÷ 232.80 = 0.3333 — exactly D ÷ c. A deposit returns the ratio of the deposit to the booking, never more. If it is a third of what a booking contributes, you recover a third of every absence you failed to prevent, and no wording in your terms moves that number.
"We got the deposit anyway" is a third of a loss, reported as a win.
Deposit, prepayment and card hold: three different promises to the guest
The arithmetic above uses one variable, D. In practice it comes from one of three arrangements, and the guest experiences them differently — so the refusal rate r differs between them even at an identical amount.
| Form | What the guest agrees to | What the venue holds | What the guest sees on cancellation |
|---|---|---|---|
| Deposit against the bill | A sum paid now, credited to the bill on arrival | Money already received | Depends entirely on your written terms |
| Prepayment for a set menu | The menu and the price are fixed in advance | Money already received, against a defined service | A cancelled service rather than a cancelled table |
| Card hold, not charged | A card is authorised, nothing is taken | Nothing yet: the right to charge under stated conditions | Usually nothing at all, if they cancel in time |
A prepayment is a different promise again: it buys a defined service, and pricing one is worked through in the catering quote article. A card hold asks the guest to trust you rather than pay you — nothing leaves their account when they book. That usually makes it the cheapest in refusals, and it makes D conditional on your actually charging it, which many rooms never do. A D you never collect is a D of zero, and the first row above is where you live.
What Polish law says, and what it does not
One distinction in Polish law is worth knowing by name. Zadatek is defined in the Civil Code: article 394 § 1 says that, absent a different contractual stipulation or custom, a zadatek given at the conclusion of a contract lets the other party withdraw without a further deadline and keep it if one side fails to perform — and demand double if it was that party who gave it (Kodeks cywilny, Dz.U. 1964 nr 16 poz. 93, consolidated text via the Sejm ELI API, read 27.08.2026). Paragraph 2 adds that on performance it counts towards the payment.
What the same code does not contain is just as useful. In that text zaliczka never appears as a defined institution, and kaucja and rezerwacja do not appear at all — zero occurrences of both, in the file read on 27.08.2026. A restaurant deposit therefore has no ready-made legal shape waiting for it, and this page does not tell you which one yours is. That is a matter for whoever drafts your terms, and worth settling before the first guest asks for their money back.
A party of ten and a table for two: why one rule cannot fit both
The rearrangement above showed that the threshold depends on D ÷ c, not on D. So if you set the deposit per person and contribution scales with the party too, the threshold does not change at all.
Check it: a two-top at 20 PLN a head gives D = 40, c = 120; a ten-top on the same basis gives D = 200, c = 600.
p and p′ both return 1 − 0.93 ÷ (0.98 + 0.02 × 0.3333) = 5.74 %.Identical.
So the reason to treat a large party differently is not in the formula. It is in its three inputs:
cis not proportional. A ten-top blocks one slot but often two servers, and a group menu rarely carries the margin of a pair ordering freely. Measure it rather than multiplying.pdiffers by party size, sometimes sharply. A large group is organised by one person on behalf of others who confirmed nothing.rdiffers most of all. A group organiser expects to be asked for a deposit — the event enquiry article walks through that conversation. A couple booking dinner for Thursday does not expect it, and some share of them will book elsewhere.
That is the case for a threshold rule rather than a blanket one. If r for groups is a fraction of r for pairs, the same terms sit above the line on ten-tops and below it on two-tops at the same time. One policy for both is not simplicity — it is two decisions, made once.
| Situation | What is at stake | A reasonable form |
|---|---|---|
| A group, a set menu, a date the room cannot resell | A whole service, ordered in advance | Prepayment for the agreed menu |
| A peak slot with a waiting list behind it | A slot you could have sold twice | A card hold, charged under written conditions |
| An ordinary table on a quiet evening | Less than the risk of losing the booking | Nothing, or a confirmation chain |
| A guest who has failed to arrive before | A repeat of a known outcome | A hold on that booking, not a house rule |
What lowers no-shows without asking for money at all
The p′ in the formula is a measurement, not a property of deposits. Anything that lowers no-shows moves it, and the cheaper instruments have a refusal rate of zero: nobody declines to book because you offered to send a reminder.
- A confirmation chain. A booking that is confirmed, reminded and trivially cancellable fails less often; the sequence is in the no-show article. A guest who can cancel in two taps frees the table instead of abandoning it, and a freed table is a slot you can resell. Running that from the booking system rather than by hand is what booking automation and automatic messages are for.
- Reaching them on the channel they used. A reminder sent where the guest already talks to you gets read; one sent elsewhere does not. That is the argument for messenger channels over a single email.
- Answering the phone at all. A booking nobody took is worse than one that failed, and invisible in the same way a refusal is; the arithmetic of that loss is in the missed calls article, and a voice assistant on the line makes the count zero when the room is busy.
- Following up afterwards. A guest who cancels politely can be rebooked; a guest who walked away over a deposit cannot, unless somebody goes back to them — which is what automated follow-up is for.
These belong in the arithmetic in a specific order: run them first, measure p again, and only then decide about a deposit. A deposit laid on top of a broken confirmation process will show a large drop in no-shows and prove nothing. The usual order of automating a restaurant — bookings, suppliers, reviews — is in the automation overview.
What the arithmetic will never show you: the guest who stops calling
The refusal rate r counts guests who declined when they booked. It does not count the guest who booked, paid, ate and quietly decided this is the restaurant that asks for money before letting you sit down, nor the friend they told. Those losses are real, slow, and arrive as a slightly weaker Thursday six months out — indistinguishable from weather.
Nothing in your data will separate that, so the threshold is an upper bound on the cost of a deposit rather than the cost itself. Three consequences follow, and they matter more than any decimal place above.
Narrow the rule as far as it will go. Every table that gets a deposit it did not need pays the invisible cost for nothing. A deposit on ten-tops on Saturdays and nowhere else carries a fraction of the exposure a house-wide rule does.
Say the condition before the guest is invested. A deposit named in the first sentence is a condition; the same deposit named after five minutes of choosing a table is a surprise, and surprises get repeated to other people. What the line should say belongs to what a system answers on the phone.
Write down what you decided and why. Judging in a year whether the deposit worked will need p from before, p′ from after and the refusal rate in between. If nobody recorded the first, it will be settled by whoever remembers most confidently — which is what analytics exists to prevent.
Frequently asked questions
Do booking deposits actually reduce no-shows?
That is a measurement, not a fact you can look up, and this page quotes no industry figure on purpose: neither the Eurostat catalogue nor the Polish statistical office carries a dataset on booking absence. The only trustworthy answer is your own — introduce the deposit on one narrow slice of your bookings and compare the no-show rate before and after. If it did not move, the arithmetic here says the deposit is not paying for a single refusal.
How many refused bookings can a deposit cost before it stops paying off?
Exactly as many as the break-even refusal rate allows: 1 − [ (1 − p) × c ] ÷ [ (1 − p′) × c + p′ × D ]. In the worked example here — 7.0 % no-shows before, 2.0 % after, 120 PLN of contribution, 40 PLN kept — the answer is 5.74 % of attempts. Change any input and the answer moves with it, which is why no published threshold would mean anything for your room.
What is the difference between a deposit, a prepayment and a card hold?
What the guest gives up at the moment of booking, and therefore how many walk away. A deposit takes money now and credits it to the bill; a prepayment buys a defined service, usually a set menu; a card hold takes nothing and reserves the right to charge under stated conditions. Only zadatek is defined in its own right in Polish law, in article 394 of the Civil Code.
Should the deposit apply to every booking or only to large parties?
The formula does not distinguish them: if the deposit scales with the party, the threshold is identical for a two-top and a ten-top. The inputs differ — a group organiser expects a deposit and a couple booking Thursday dinner does not. That alone usually puts a deposit above the line for groups and below it for pairs at once, which makes a narrow rule the safer default.
Does the amount I keep equal the money the no-show cost me?
No, and the shortfall is exact: you recover the ratio of the deposit to the booking, D ÷ c, and nothing more. In the example here a 40 PLN deposit against a 120 PLN contribution returns a third of every absence it failed to prevent — 77.60 PLN against 232.80 PLN lost. Reporting kept deposits as a saving counts a third of a loss as a win.
What reduces no-shows without asking for money upfront?
Confirming the booking, reminding the guest on a channel they actually read, and making cancellation trivially easy so the table comes back instead of standing empty. All three have a refusal rate of zero, so they never appear on the wrong side of this trade. Run them first and measure the no-show rate again.
Calculate your own break-even refusal rate from those four numbers before you introduce a deposit anywhere — and if it lands near zero, the honest conclusion is that your problem is not the absence of a deposit but the absence of a confirmation. The rest of the series lives in the restaurant section.