Restaurant acquisition cost is the money a restaurant spends on attracting guests in a period, divided by the guests it newly recognised in that same period. It comes in two versions — blended and paid — that use different denominators and answer different things for the owner. Payback on that cost is best expressed in guest visits.
What counts as acquisition spend and what belongs to keeping a guest
Acquisition spend — money spent to bring in a guest who has not been here before: advertising, listing and portal fees, first-visit offers, and the labour behind all of it.
"Marketing" is not a category an accountant can settle. A loyalty top-up and a paid post leave the bank account the same way and both get filed under promotion. But one buys a guest you do not have yet, the other pays to keep one you already have. Put them in the same numerator and the number answers neither of the two things you wanted to know.
So the first job here is boring and unavoidable: sort your own spending into three buckets before you divide anything.
| Spend | Acquisition | Retention | Neither |
|---|---|---|---|
| Paid ads aimed at people who have never been here | yes | — | — |
| Listing or portal fee that brings first-time bookings | yes | — | — |
| Discount on a first visit, counted as margin given away | yes | — | — |
| The hours your team spends making and running campaigns | yes | — | — |
| Loyalty points earned by regulars | — | yes | — |
| Birthday offer sent to the guest list | — | yes | — |
| Reminder message before a booking | — | yes | — |
| Photography reused across everything for a year | — | — | split it, or leave it out and say so |
Two lines of it are worth arguing about, once.
A first-visit discount is spend, not a discount. A guest who walks in with a voucher leaves you less than the full contribution of that visit. The margin you gave away is the price of getting them through the door, and it belongs in the numerator exactly like an invoice from an ad platform does. Leave it out and acquisition looks cheap for the arithmetic reason that part of its cost moved into a line called "revenue that did not arrive".
Labour is spend too. The evening your manager spends rewriting the campaign is paid time. If you run promotion in-house, most of your true acquisition cost sits in a payroll line, and the version of the number that ignores it is the version that flatters you. The same holds for the tooling around the funnel: the cost of a system that collects and answers inquiries is part of what a new guest costs, not a separate story about overhead.
The rule to write down before you count
Whatever you decide, write the decision down and date it. The formula is not what breaks; the boundary is. A number produced under one rule in March and another in June is two numbers wearing one name, and the first thing anybody does with them is subtract one from the other.
Who a new guest is when the guest has to be recognised first
Newly identified guest — a guest recognised by the restaurant for the first time inside the period.
That is the whole difficulty of this page in one line. A shop knows its customer because the customer logs in. A restaurant usually does not: somebody sat down, ate, paid by card and left, and nothing in that sequence tells the till whether they had been here before.
The definition of recognition is not owned by this page. It belongs to the page on returning guests and how to count them, which defines an identified guest as a guest whose two visits can be linked by the same identifier: a booking, a loyalty account, a delivery address, a card token. Take it from there unchanged — the moment two of your pages define recognition differently, their numbers stop being comparable inside your own restaurant.
What follows is worth saying plainly: your acquisition cost is not the cost of a new guest. It is the cost of a new guest you are able to see. If a third of your checks carry an identifier, the denominator counts a third of the people who came for the first time, and the cost per guest comes out correspondingly higher than the truth. That is not a reason to abandon the number — it is a reason to print the identification rate next to it, in the same row of the same report, every time.
Two practical consequences:
- raising the share of checks you can recognise moves this number without a single unit of spend changing, so a fall in acquisition cost after you install a guest card layer may be recognition improving rather than marketing improving;
- the guests you recognise are not a random sample of the room. People who book, order delivery or sign up for a card select themselves, and they tend to be the returning kind. The layer that stores them — a guest database or the heavier system layer you add next — decides who is visible to every formula here.
Blended acquisition cost: the formula and what it actually measures
Blended acquisition cost — all acquisition spend of a period divided by all guests newly identified in that period, attributed to nothing.
Blended acquisition cost = All acquisition spend in the period ÷ Newly identified guests in the period
All acquisition spend in the period— everything from the acquisition column of your own table above: paid media, listing fees, margin given away on first visits and the labour behind them, PLN;Newly identified guests in the period— distinct guests recognised for the first time inside the same period, people;- the result is PLN per guest.
Take one calendar month as the worked example of this page. Acquisition spend came to 28,200 PLN: 14,400 on paid media, 3,600 on portal and listing fees, 5,700 of contribution given away on first-visit offers and 4,500 of internal labour. In the same month 470 guests were recognised for the first time.
28,200 ÷ 470 = 60.00 PLN per newly identified guest
Check it backwards: 470 × 60.00 = 28,200 PLN. The unit matters more than the value — this is money per guest, not per visit and not per booking. A restaurant quoting its acquisition cost without saying which of the three it divided by has quoted nothing.
The blended version is good for the total: for every new face this restaurant learned to recognise last month, this much money left the account. It does not know which channel did it and does not pretend to — which is why it is the only version nobody can inflate with a flattering attribution rule.
The degenerate case that has to be printed, not hidden
If no guest was newly identified in the period, this formula has no value. Not zero, not infinity — no value, because you cannot divide by a count of nobody. A spreadsheet showing a dash there tells the truth; one showing a large round number tells you it substituted something. The same holds one step earlier: if your identification rate is zero, every number here is undefined, and the work is not marketing but recognition.
Paid acquisition cost: a different formula and a different denominator
Paid acquisition cost — paid acquisition spend divided by the newly identified guests attributed to it by a written rule.
Paid acquisition cost = Paid acquisition spend ÷ Newly identified guests attributed to paid channels
Paid acquisition spend— only money paid to channels you can switch off: media buying, portal and listing fees, PLN;Newly identified guests attributed to paid channels— the guests your written attribution rule assigns there, people;- the result is PLN per guest: the same unit as the blended version, a different population.
In the same month, paid spend was 14,400 PLN and the attribution rule assigned 180 newly identified guests to it.
14,400 ÷ 180 = 80.00 PLN per newly identified guest
Check it backwards: 180 × 80.00 = 14,400 PLN.
Now put the two next to each other and read what they say.
| Version | Numerator | Denominator | What it is for |
|---|---|---|---|
| Blended | all acquisition spend of the period | all guests newly identified in the period | the size of the whole bill: what a new face costs on average |
| Paid | only spend on channels you can switch off | only guests a written rule assigns to those channels | one decision: keep, cut or raise the paid budget |
Sixty and eighty are both correct, and they are not two estimates of one quantity — they are two quantities. The blended number is diluted by everybody who walked in from the street, came with a neighbour or found you on a map without a paid click; the paid number is not. Expecting them to converge is expecting a share of a room to equal a share of a queue.
The rule matters more than its precision
No attribution rule is true. Last click, first click, a coupon code, "the guest said so at the table" — each is a convention, and each will be wrong about individual guests. That is survivable. Having no written rule is not, because then the rule changes silently: a new person joins, counts differently, and the number moves without anybody touching the marketing. Keeping the two versions apart is measurement work, and it is what a proper analytics layer is for.
Why these two numbers cannot be compared between restaurants
This is the part that gets skipped, so it is not a footnote here.
Do not compare your acquisition cost with another restaurant's — not with a competitor's, not with a figure on a conference slide, not with a number a supplier quotes at you. The formulas above are divisions whose denominators each restaurant defines for itself, and there are at least four independent places where two honest restaurants will disagree:
- What went into the numerator. One counts the margin given away on first visits, the other does not. One counts the manager's hours, the other treats them as free.
- Who is in the denominator. Recognition depends on how many checks carry an identifier, and that share differs several-fold between a bookings-heavy dining room and a walk-in bar.
- Which version was quoted. Sixty against eighty in this example — and nothing in the phrase "our acquisition cost" says which of the two you are hearing.
- What the guest is worth there. A cost is only high or low against the value of what it buys, and that value differs per restaurant more than the cost does.
The legitimate comparison is with your own previous period, counted under the same written rule. It is also the comparison that changes decisions, because it is the only one where everything except your own marketing is held still.
There is no official benchmark to appeal to instead, and this is not modesty. Neither the European statistical office nor the Polish one publishes marketing spend of restaurants as a separate line. The structural business statistics dataset for food and beverage service activities publishes forty-eight indicators for Poland — enterprises, persons employed, wages, employee benefits expense, social security costs, value added, gross operating surplus, purchases of goods and services, purchases of energy, expenses on agency workers, gross investment, net turnover and ratios built from those (Eurostat, sbs_ovw_act, NACE I56, Poland, dissemination API, opened 27.08.2026) — and not one of the forty-eight is advertising, marketing or acquisition. A number found in neither the national nor the European official statistics is one you build yourself or do without.
That is why both versions belong on one screen rather than in two conversations: side by side they cannot be mistaken for each other, quoted separately they will be. Putting them in one place is what a shared reporting surface is for, and which numbers an owner actually looks at decides whether both survive a Monday morning.
Payback measured in visits rather than in months
Payback in visits — how many visits by one guest it takes for their accumulated contribution margin to cover what was paid to acquire them.
Payback in visits = Acquisition cost ÷ Contribution margin per visit
Acquisition cost— either version above, PLN per guest, and you say aloud which one you used;Contribution margin per visit— the contribution a single visit leaves behind, PLN per visit;- the result is visits: PLN divided by PLN per visit leaves visits.
The contribution figure is not defined here. It is defined on the break-even page, which states that contribution margin is the money left from net sales after variable costs, available to cover fixed costs and, once they are covered, to become profit (restaurant break-even point). Per visit, that is what one guest leaves after the food, packaging, platform commission and card fee of that visit are gone.
With a contribution margin of 24.00 PLN per visit:
- blended:
60.00 ÷ 24.00 = 2.5 visits; - paid:
80.00 ÷ 24.00 = 3.33 visits.
Check the first backwards: 2.5 × 24.00 = 60.00 PLN.
Visits do not come in halves
A guest cannot make a third of a visit. Two and a half visits means the money comes back inside the third one; 3.33 means the fourth visit clears it. Round up when you turn this into a decision, and round nothing when you record the number.
Why not months
Because a month is not a property of the guest. To turn visits into months you must assume how often that guest comes back — the one thing you have not measured. That assumption lives on the returning guests page and differs by segment, season and neighbourhood. A restaurant sells no subscription; nothing obliges a guest to arrive on a schedule. Visits keep the sentence to what was counted, and turn payback into an operational matter: what has to happen for the second and third visits to exist.
The degenerate case
If contribution margin per visit is zero or negative, there is no payback at any number of visits: each visit adds nothing or takes something away. When that happens the work is on the margin of the visit, not on the marketing — and the arithmetic says so before the marketing has a chance to be blamed.
Attribution: the guest came because of the ad, or would have come anyway
The paid version divides by guests "attributed to paid channels", and attribution has a weakness no rule fixes: it records where a guest was last seen, not why they came. A regular who searched for your name, clicked a paid link sitting above your own listing and walked in would have walked in anyway. Your attribution rule counted them; your marketing did not create them.
Separating the two needs a different method — a holdout group, a district left untouched, a week left un-promoted. It belongs to the page on telling a working promotion from a good week. What this page owes you is the warning: your paid acquisition cost is an upper bound on how many guests the money created, and a bound is not a measurement.
Two ordinary things push that bound further from the truth:
- inquiries that never became guests. Money was spent, the person raised a hand, nobody came back to them. They sit in the numerator and not in the denominator, which makes acquisition look expensive for a reason unrelated to the channel. What happens to an inquiry after the first conversation moves this number more than a change of creative;
- guests who never raised a hand at all. The reasons people do not leave an inquiry sit above every formula here, and a form that actually collects them changes both halves of the fraction at once.
Acquiring a new guest against winning back a lapsed one
These are different budgets, and mixing them is the second most common way this number is spoiled, after mixing the two versions.
A lapsed guest is already recognised: you know what they ordered, when they last came and what a visit of theirs is worth. Money spent on them is retention spend. It belongs in the other column of the table at the top of this page and must not enter either numerator here, or the denominator — which counts only guests recognised for the first time — no longer matches the money above it.
Comparing the two budgets is legitimate as long as each side is calculated on its own terms: acquiring a new guest buys a relationship with no history, winning back a lapsed one costs what a message costs and buys a relationship whose value you can already estimate. Which deserves the next unit of money is a decision, not a formula — but a decision that needs both numbers to exist.
What this number does not say: it is not about the quality of the guest
An acquisition cost is a price. It says nothing about what you bought: a guest who returns nine times and one who never comes back cost exactly the same to acquire, so two restaurants at the same 60.00 PLN per guest can be in opposite situations. The number becomes meaningful only against a ceiling, and that ceiling is the value of the guest over their whole time with you — a separate page and a separate calculation.
Acquisition ceiling test: Acquisition cost < Affordable acquisition cost
Acquisition cost— the version you chose, PLN per guest;Affordable acquisition cost— the lifetime contribution of a guest multiplied by the share of it the owner is prepared to spend, PLN per guest, taken from the lifetime value page;- both sides are PLN per guest, so the comparison is legal.
Suppose the lifetime value page gave 420.00 PLN of lifetime contribution per guest and the owner decided to spend at most fifteen percent of it on acquiring one. The affordable cost is 420.00 × 0.15 = 63.00 PLN, and the two versions of the same month give opposite verdicts:
- blended 60.00 < 63.00 — inside the ceiling;
- paid 80.00 > 63.00 — outside it.
Nothing is wrong with the arithmetic. This is what it looks like when the average is affordable and the marginal, bought guest is not, and it is the clearest demonstration of why the two versions are not interchangeable. The owner who quotes the blended number keeps the budget, the owner who quotes the paid number cuts it, and both are quoting "our acquisition cost".
Three more things this number cannot tell you, so that nobody asks it to:
- whether the money produced the guest. That is the holdout matter, answered on its own page;
- whether the channel earned its keep in percentage terms. Return on advertising spend is measured on margin, not revenue, and it is its own calculation with its own trap: that page returns a ratio, this one a price;
- what a booking costs. A booking is not a guest: one guest can make several bookings, one booking can bring six guests, and a no-show consumes a booking without producing a guest at all. Cost per booking is money per booking and lives on the cost-per-booking page; this page is money per guest. Same currency, different units, and dividing one by the other produces nothing.
Where a machine does the counting here and where it is plain arithmetic
Both formulas here are division. There is no model in them, no estimate and nothing to train — you can run them on paper. If a supplier says their software calculates acquisition cost with artificial intelligence, what it does is the division above on data it already has.
What genuinely benefits from a model sits next to the formulas, not inside them: recognising that two visits belong to one person when the identifiers do not match exactly, and sorting inquiries by how likely they are to become guests before somebody spends an evening on them. Both change the denominator. Neither makes the cost per guest more accurate; they make the count of guests more complete, which is a different and more useful thing.
Aura is a management layer over the numbers a restaurant already produces. It does not run your advertising and does not promise you a guest.
Frequently asked questions
How do I calculate customer acquisition cost for a restaurant?
Add up everything you spent in one period on bringing in people who had never been to you — paid media, listing and portal fees, the margin given away on first-visit offers and the labour behind all of it — and divide it by the guests you recognised for the first time in that same period. The result is money per guest. Write down which lines you included: the rule matters more than the value.
What is the difference between blended and paid acquisition cost?
Both the numerator and the denominator change. Blended takes all acquisition spend over all newly identified guests, attributed to nothing; paid takes only spend on channels you can switch off, over only the guests a written rule assigns there. The worked example on this page gives 60.00 PLN blended and 80.00 PLN paid for one month. Both are correct, they measure different populations, and neither can be compared with another restaurant's figure.
Who counts as a new guest if most guests are anonymous?
A guest you recognised for the first time — recognised meaning two of their visits could be linked by the same identifier: a booking, a loyalty account, a delivery address or a card token. Guests you cannot recognise are not in the denominator, which is why the identification rate has to be printed next to the cost. Recognise a third of your first-time guests and your cost per guest is roughly three times what it would be if you recognised all of them.
Why is payback expressed in visits and not in months?
Because visits are what you count and months are what you would have to assume. Turning visits into months requires a figure for how often a guest returns, which varies by segment and season and is measured on a different page. Visits keep the answer inside what was observed, and they turn payback into an operational matter: what has to happen for the second and third visit to exist.
How much may I spend to acquire one guest?
Up to the share of a guest's lifetime contribution you have decided in advance to spend on acquiring them. That ceiling comes from the lifetime value calculation, not from this page and not from an industry figure — there is no published benchmark for restaurant acquisition cost in the Polish or the European official statistics. The test compares two numbers in the same unit: cost per guest against affordable cost per guest.
Does a first-visit discount belong in acquisition cost?
Yes, as the contribution you gave away rather than as the face value of the voucher. A guest arriving with a first-visit offer leaves less margin behind than one paying full price, and that difference is what the visit cost you to buy. Leaving it out does not make acquisition cheaper; it moves part of the cost into revenue that never arrived, where nobody will look for it.
Count both versions for the same month and put them side by side, with the identification rate next to them. The gap between them is exactly the share of new guests you credit to your marketing without having earned the credit — and the ceiling that says whether either number is affordable waits one page over, on the restaurant hub.