The cost of one departure is the sum of four blocks: separation, hiring, onboarding and the productivity gap while the position is empty or held by someone not yet up to speed. The last block is the largest and the only one that never arrives as an invoice, which is why owners routinely underestimate what turnover costs them.
Why a departure never shows up as a line in your profit and loss
Open your accounts for the month a cook walked out and try to find the event. You will not. The final settlement hides inside payroll, where it looks like an ordinary wage. The job advert hides inside marketing or inside "other". The manager's evening spent reading applications hides inside that manager's salary, which you were going to pay anyway. And the largest item — the food that came out slower, the covers you did not turn, the dishes the trainee had to remake — never enters the books at all, because nobody invoices you for output that did not happen.
That is the whole trick of turnover. It is not an expense you approve; it is a number you assemble afterwards from four places, only one of which your accountant already knows about. This page assembles it, on your own figures, and refuses to hand you somebody else's benchmark instead.
Every statutory figure quoted below is Polish: the notice period, the paid job-search leave, the holiday equivalent. Each one comes from the Labour Code as published in the Journal of Laws, with a link you can open yourself.
What Polish statistics publish about this, and what is simply not there
There is no official turnover rate for Polish restaurants. That is not a shrug, it is a search with a result. The Eurostat catalogue of datasets contains no labour turnover, separation or hiring-rate series at all; what it does carry next to this question is three different things — how long people have been in their current job (lfsa_egdn2), how many posts stand vacant (jvs_a_rate_r2), and how many people work in the sector (lfsa_egan2). The Polish statistical office organises its labour-market section into seven subsections — methodology, registered unemployment, employment and wages, the labour force survey, working conditions, publications, and labour demand. Staff turnover is not one of them, and the words for it do not appear on the page.
So this page names no "normal turnover" and no "replacement costs N salaries" rule. Both circulate widely and neither has a first source behind it. What the official numbers do give you is the shape of the sector you are hiring into:
| What is published | Poland, section I, 2025 | What it does not tell you |
|---|---|---|
| Average monthly gross wage in accommodation and food service | 5 990.92 PLN — the lowest of all PKD sections, 32.7 % below the national average of 8 903.56 PLN (GUS, Przeciętne zatrudnienie i wynagrodzenia w gospodarce narodowej w 2025 r. — dane wstępne, published 18.06.2026) | What replacing one person costs you. A wage is what you pay the position, not what losing it costs |
| Employed persons, employees, and employees on fixed-term contracts | 395.6 thousand employed (Eurostat, lfsa_egan2), of them 337.1 thousand employees (lfsa_eegan2) and 129.2 thousand on fixed-term contracts (lfsa_etgan2) — Poland, NACE section I, ages 15–64, data updated 30.06.2026 | How often those contracts end. A stock of people is not a flow of departures |
| Employed in their current job for less than three months | 22.6 thousand out of 395.6 thousand (Eurostat, lfsa_egdn2, Poland, section I, ages 15–64, 2025) — 5.7 % on our own division of the two published figures | Whose leaving created those vacancies, and at what cost |
| Job vacancy rate | 1.0 % on annual average, and 0.6 % among enterprises with ten or more employees (Eurostat, jvs_a_rate_r2, Poland, section I, 2025, data updated 20.03.2026) | How long your post stays open. A sector average is not your hiring queue |
Two caveats belong right here, next to the numbers rather than in a footnote. First, all of the above is NACE section I, "accommodation and food service activities" — hotels are inside it and there is no separate cut for restaurants alone. Writing "Polish restaurants" next to these figures would be a lie, so we do not.
The four blocks: separation, hiring, onboarding, productivity gap
Separation cost — everything the ending of employment costs: final settlement, handover time, administrative work.
Hiring cost — advertising the role, the hours spent screening and interviewing, valued at the fully loaded hourly cost of whoever spent them.
Onboarding cost — training hours of the newcomer and of the people training them, plus the wastage of the learning period.
Productivity gap — the output lost between the departure and the moment the replacement reaches normal speed. Measured in hours multiplied by the value the position produces, not by the wage it earns.
| Block | What sits inside it | How you measure it | Where it hides in the books |
|---|---|---|---|
| Separation | Final settlement, holiday equivalent, handover hours, paid job-search leave | Hours × fully loaded hourly cost, plus statutory payments | Payroll — indistinguishable from an ordinary month |
| Hiring | Job advert, screening, interviews, trial shifts | Spend, plus manager hours × their fully loaded hourly cost | Marketing, or nowhere |
| Onboarding | Trainer hours, wastage of the learning period | Hours × fully loaded hourly cost, plus written-off product | Wages and food cost |
| Productivity gap | Output the position did not produce | Hours × share missing × value produced per hour | Nowhere at all |
Separation: what Polish law puts on the clock
Three of the separation items are not a matter of your judgement, they are set by statute. The notice period for a contract of employment depends on length of service with you and is two weeks under six months, one month from six months, three months from three years (Labour Code, art. 36 § 1, consolidated text Dz.U. 2025 poz. 277). When it is you who gives notice and the notice runs at least two weeks, the employee is entitled to paid time off to look for work — two working days on a two-week or one-month notice, three working days on a three-month one (art. 37 § 1–2, same consolidated text). And any holiday left unused when the employment ends must be paid out in cash (art. 171 § 1). All three carry your employer contributions on top, so price them at the fully loaded hourly cost and not at the nominal rate.
The rate itself belongs to another page, and the difference between the two is worth naming out loud before you use it.
Fully loaded hourly cost — what one hour of a person costs the employer once everything sitting on top of the gross wage is counted, not the rate on the payslip. The term belongs to our labour-cost pillar, which defines it in a single line — Fully loaded hourly cost = Total annual cost of the position to the employer ÷ Hours actually worked — and walks the whole build-up of contributions through to ≈ 39.08 PLN/h on its worked example: restaurant labour cost as a percentage. The difference is the unit, and the unit is the whole boundary between the two pages: there the answer comes out in PLN per hour, here in PLN per replacement. You take their hourly figure as an input; you do not recount it.
The productivity gap: the biggest part and the invisible one
The gap starts on the last shift of the person leaving, not on the day the post is filled. Between those two dates the position produces less than it should — first because nobody is in it, then because someone is in it who is still asking where things are.
Productivity gap = Hours below normal output × Share of output missing × Value produced per hour by the position
Hours below normal output— from the departure to the moment the replacement reaches normal speed, in hours;Share of output missing— the average shortfall over that stretch, a dimensionless fraction from 0 to 1;Value produced per hour by the position— what the position produces in an hour, in PLN per hour.
The third variable is where people go wrong, and always in the same direction. Substituting the wage understates the loss by exactly the contribution the position makes: a cook who costs you forty PLN an hour and produces a hundred and twenty is a ninety-PLN hole when the station is short, not a forty-PLN one. Sales per labour hour is the ready-made way to get that value out of your own till — the method sits in sales per labour hour, and the position-level share of it is your own reading of the rota.
Where the hours come from if nobody writes them down
Most kitchens do not record the ramp-up, so the first count is a reconstruction: how many shifts ran short, how many were covered by overtime, and on what date the head chef stopped correcting the new person's plates. One caveat that changes the answer: if the vacancy was covered by overtime rather than left empty, the gap moves rather than disappears. Output held up, so the third block shrinks — but the hours were bought at a premium rate, and that premium belongs in your labour cost. Counting both at full weight is double counting; counting neither is how the whole thing became invisible in the first place. Keeping shift and task history where it can be read back later is exactly what team management and task handover are for.
The formula for one replacement
Cost of one replacement = Separation cost + Hiring cost + Onboarding cost + Productivity gap
All four are money, for one position, over one departure. Here is the arithmetic on placeholder numbers — the figures are stand-ins so the shape is visible, and the only real thing in the example is the order of magnitude between the blocks. Substitute your own.
| Line | Arithmetic | Result |
|---|---|---|
| Handover, manager's time | 6 h × 55.00 PLN/h | 330.00 PLN |
| Paid job-search leave, 2 working days | 16 h × 39.08 PLN/h | 625.28 PLN |
| Holiday equivalent, 5 days | 40 h × 39.08 PLN/h | 1 563.20 PLN |
| Separation cost | 2 518.48 PLN | |
| Job advert | 300.00 PLN | |
| Screening and interviews | 9 h × 55.00 PLN/h | 495.00 PLN |
| Hiring cost | 795.00 PLN | |
| Trainer's hours | 14 h × 55.00 PLN/h | 770.00 PLN |
| Wastage of the learning period | 260.00 PLN | |
| Onboarding cost | 1 030.00 PLN | |
| Productivity gap | 320 h × 0.35 × 120.00 PLN/h | 13 440.00 PLN |
| Cost of one replacement | sum of the four blocks | 17 783.48 PLN |
The 39.08 PLN per hour is not invented either: it is the fully loaded hourly cost worked out for a full-time position at the statutory minimum on our labour cost page. The manager's 55.00 PLN and the 120.00 PLN of value per hour are placeholders and yours will differ.
The gap alone is 75.6 % of the total. Set the whole number against the average monthly gross wage in the sector and it comes to about three times it — which is the reason "a departure costs a month of wages" survives as folklore and fails as arithmetic.
The double count that ruins this sum
There is one place where this formula quietly counts the same money twice: the newcomer's own hours during training. They are wages you pay, so they feel like an onboarding cost; they are also hours of missing output, so they are already inside the productivity gap. Put them in both and your total is wrong by their full value. The rule that keeps it clean is the one used above — onboarding holds only the trainer's hours and the wastage; the newcomer's under-performance lives in the gap and nowhere else.
Turnover rate: the formula, the denominator and the period
Turnover rate — departures in a period divided by the average number of people employed in that period, read as a percentage of that period and of no other.
Turnover rate % = Departures in the period ÷ Average headcount in the period × 100
Departures in the period— people who left, in persons;Average headcount in the period— the average number employed across the same period, in persons — the average, not the figure on the closing day.
Persons divided by persons leaves a percentage that belongs to its period and to nothing else.
The same seven over a month would be a different, much larger-looking number, and putting the two side by side means nothing at all. Any turnover figure without its period attached is unreadable, including your own from last year.
The closing-day headcount is the more common mistake, and it flatters you in exactly the wrong season: staff up for summer, lose people in October, count against the October number, and the rate reads lower than the year deserved.
Annual turnover cost and its share of your payroll
Annual turnover cost = Departures per year × Average cost of one replacement
Departures per year— departures over twelve months, in departures;Average cost of one replacement— weighted by position, in PLN, because a kitchen and a floor replacement are not the same event.
Continuing the example: three kitchen departures at 17 783.48 PLN and four floor departures at 8 900.00 PLN give 88 950.44 PLN a year, and a weighted average of 12 707.21 PLN per replacement.
It is not a benchmark and we are not offering it as one; it is what those inputs produce, and your own inputs will produce something else. Where a number like this belongs on a screen, next to the rate rather than buried in an annual report, is covered in which numbers an owner actually looks at and set up in dashboards.
Why kitchen and floor have to be counted separately
Averaging the two positions destroys the answer, because almost every variable differs.
| Position | Ramp-up length | Whose time the training eats | How the vacancy gets covered |
|---|---|---|---|
| Line cook | Long — stations, recipes, timing under pressure | Head chef, at the busiest hours | Overtime of the remaining kitchen, or a shortened menu |
| Head chef | Longest — suppliers, costing, the whole menu | Owner, plus every station | Rarely covered at all |
| Waiter | Short for the floor, long for the menu | Shift leader, during service | The rest of the shift takes more tables |
| Bar | Medium — speed matters more than range | Bar lead | Queue at the bar, slower table turns |
Two consequences follow. A kitchen departure carries a longer gap at a higher value per hour, so it can cost several times a floor departure even though both people earned similar money. And the cover mechanism decides where the cost lands: a shortened menu shows up as lost sales, a covered shift shows up as labour cost. Both are real; they are simply in different rows.
What disappears from these costs when someone leaves at the end of a season
Timing changes the arithmetic more than most owners expect, and it changes it in one block only. If the person leaves as the season closes and the position is genuinely not needed until spring, the productivity gap for the closed weeks is zero — there is no output to miss. Separation and hiring do not move: the notice period, the holiday equivalent and the advert cost the same in November as in June.
What does not disappear is the onboarding, and it usually gets worse. Hiring in the low season means the newcomer learns on quiet shifts and then meets the first busy weekend as a beginner anyway, so the ramp-up stretches across the boundary instead of ending inside it. Count the gap against the hours the position was actually needed, not against the calendar, and the seasonal case stops looking like a discount.
Retention spend against replacement cost: the comparison that is legitimate
This is the comparison the whole page exists to make possible, and it has a formula:
Maximum justified annual retention spend = Cost of one replacement × Probability that the spend prevents the departure within the year
Cost of one replacement— the four blocks, in PLN;Probability that the spend prevents the departure— a dimensionless fraction from 0 to 1.
Money multiplied by a dimensionless fraction is money, and the answer is a ceiling on the year's spend, not a recommendation. On the example above, at a one-in-four chance of preventing the departure, the ceiling is 4 445.87 PLN a year for that position. A raise of 2.00 PLN per fully loaded hour across 1 800 hours costs 3 600.00 PLN a year and therefore sits under the ceiling.
The degenerate case is worth printing rather than hiding: if the probability is zero, the justified spend is zero. Money that changes nobody's mind is not retention, it is a raise — which may be right for other reasons, but not for this one. And the probability is a judgement you make, not a measurement anyone can hand you; the formula is honest precisely because it puts your guess in a visible place instead of burying it.
What this arithmetic does not decide
It does not tell you who to keep. A replacement cost of eighteen thousand PLN says the event is expensive; it says nothing about whether that particular person was worth eighteen thousand PLN to keep. Those are different questions and the second one is not arithmetic.
It also does not tell you the cause. High turnover with a high cost per departure and high turnover with a low one are different illnesses — the first is about the positions, the second is about the pipeline — and the sum alone cannot separate them. And there is no artificial intelligence anywhere on this page: everything above is addition, multiplication and one division, done on your figures. Where a system helps is upstream of the arithmetic — holding the shift history, the task handovers and the vacancy dates so that next year's count takes an hour instead of a week, which is the practical half of what can actually be handed to a system and of the four thresholds where manual work stops paying. Whether that knowledge sits in a system or in the departing person's head is the same fork described in CRM or ERP, and the everyday version of it in restaurant automation.
Where turnover sits among the other numbers you watch is a separate decision, laid out in the restaurant KPI guide; where labour sits inside your cost structure at all is prime cost. The same logic of counting person-minutes rather than events runs through how many calls you can actually take. Pulling the four blocks into one figure is a job for the finance layer, and counting how many replacements a year you actually had is a job for analytics.
Frequently asked questions
How do I calculate the cost of one employee leaving?
Add four blocks: separation, hiring, onboarding and the productivity gap. Take the separation items from statute and payroll, the hiring items from your spend plus the manager hours, the onboarding from the trainer's hours plus the wastage, and the gap from hours below normal output multiplied by the share missing multiplied by the value the position produces per hour. Price every hour at fully loaded cost, not at the nominal rate.
Why is the productivity gap the biggest part of the cost?
Because it runs for weeks while the other three are single events, and because it is valued at what the position produces rather than at what it is paid. In the worked example above it came to three quarters of the total. It is also the only block with no invoice behind it, which is why it is the one that gets left out.
What is a normal turnover rate in hospitality?
We do not give one, because no official statistics office publishes it in a cut that would apply to a single restaurant. The Eurostat catalogue holds no labour turnover series at all, and the Polish office's labour-market section has no such subsection. Any number you see quoted as an industry norm has no first source you can open. Compare yourself with yourself a year earlier instead — same denominator, same period, same positions.
Should I count turnover monthly or annually?
Annually for the headline, monthly only inside a year you are investigating. The two are different numbers and cannot be compared with each other. Whichever you pick, keep the denominator as the average headcount over that same period rather than the number on the closing day, and write the period next to the figure every time.
Do seasonal departures cost the same as year-round ones?
No, and the difference sits in one block. If the position is genuinely not needed while the season is closed, the productivity gap for those weeks is zero. Separation and hiring are unchanged. Onboarding often gets worse, because a person hired into quiet shifts still meets the first busy weekend as a beginner.
How do I compare the cost of keeping someone with the cost of replacing them?
Multiply the replacement cost by your honest estimate of the chance that the spend actually prevents the departure this year. That product is the ceiling on what the retention is worth. Spend under the ceiling and the comparison holds; spend over it and you are buying something other than retention.
Count your last replacement through the four blocks before you decide anything about pay. The number tends to come out large enough that the conversation about keeping people starts on its own — and once you have it for one position, the annual figure is a multiplication away. The method, the formulas and the two caveats above are all you need; the restaurant hub has the rest of the economics that sit around it.