In this article
Ask an owner who covers for the best server tonight if she calls in sick, and the answer comes without hesitation: someone steps in, service runs a little slower. Ask the same question about the person who is the only one who knows the espresso machine, can talk a guest through the wine list, or can close the till at the end of the night, and the answer changes. That evening the venue does not run slower on that one point — it does not run at all.
Every venue has a rota. Almost none has a list of which of those jobs can only be done by one person. That is exactly why this site's free staff skills matrix tool exists: it lines up all 22 jobs a venue runs on — spread across the kitchen, the bar, the floor and the building — and shows which of them sit on exactly one pair of hands. What that tool does not do is explain why that number matters so much. That is what this guide does.
It is not a staffing question, it is arithmetic. Every job only one person can do is a separate risk that can fire on any given night — a flu, a flat tyre, a family emergency. Stack a few of those risks together and the chance that at least one of them goes off tonight climbs faster than most owners would intuitively guess. That is exactly why insurers and reliability engineers use the same arithmetic: not the chance that one specific thing fails, but the chance that something, somewhere in the chain, does.
This guide works through seven numbers that map that risk: how many of these jobs a small independent venue typically carries, how likely it is that one goes uncovered tonight, what a dark job actually costs that evening, how long cross-training takes against hiring a replacement, when that cross-training pays for itself, what happens when the one person who holds the skill leaves, and how much cover you actually need before the risk practically disappears. At the bottom, run your own venue through the calculator.
Everything runs in your own browser: nothing is sent anywhere and nothing is stored. The numbers in this guide are either lifted directly from earlier guides on this site — the sick-leave figures, the replacement costs from the turnover guide — or an explicitly labelled worked example, never an invented statistic.
Why nobody ever writes that list down
A rota answers one question: who is working, and when. It never answers the question that actually matters on the night someone cannot make it in: who can stand in for them? On paper, a five-person shift looks fully staffed. It is only once someone is missing that it turns out three of those five jobs can be picked up by anyone on the floor, and two of them by exactly one person — and that difference appears nowhere on the rota.
The reason is not carelessness, it is that nobody ever asks the question out loud. A kitchen organises itself around stations and shifts the way the classic brigade has for a century — who stands where, not who could stand there if the first choice is out. That second map usually exists only in the owner's head, and even there mostly as a feeling: "if Marie is out, we have a problem" is a feeling, not a list.
And because the problem is invisible until the moment it happens, it almost never feels like something worth fixing today. That is exactly why a number does more work here than a warning: it turns a risk that always feels "probably fine" into a percentage concrete enough to actually clear an evening in the diary for.
The ultimate guide Managing Staff: the Complete Guide From rota to retention: everything that keeps your team running, in one place. Read the guideThe 7 numbers, and what each one tells you
They run in the order they actually hit a shift: first how much risk your venue is carrying, then how often it fires, what it costs, what the fix costs, and finally how much cover you genuinely need.
1. 22 jobs, and a team of five will already have five with only one name next to them
This site's free staff skills matrix tool counts 22 concrete jobs that keep a hospitality venue running, spread across four areas: the kitchen, the bar, the floor and the building itself — from "hot kitchen" and "dishwashing" to "closing the till" and "taking in deliveries". None of them is optional, and none of them is automatically covered by more than one person.
Work it through for a team of five: splitting 22 jobs over five people means each of them has more than four jobs against their name on average. Double coverage — a second person who can also do the same job — costs training time a small team rarely schedules in advance. The result is predictable: the everyday jobs (hot kitchen, dishwashing) end up covered by several people simply because everyone pitches in at some point, while the specific jobs — the espresso machine, the wine list, bookings, the allergen chart, closing the till — quickly settle on exactly one name.
The worked example below makes that concrete: five jobs with exactly one name next to them, against two jobs covered by several people. That is not a badly run venue — it is what happens in almost every small team, until somebody writes it down.
Seven jobs from the worked example above — five of them have only one name on the list.
Hot kitchen and dishwashing look "just busy" on any rota — but in practice they are the jobs everyone ends up covering. The five amber jobs are not: nobody ever decided only one person could run them, it simply was never organised any other way.
2. For five of those jobs, the chance that at least one falls through tonight is 18.5%
Working that out needs one extra number: how often someone is absent on a given day. That figure is not invented — it already sits in this site's own guide to staff absence: a team that logs 9 sick days per full-time employee per year, divided by the 225 working days a full-time employee typically works in a year, comes out at a 4% chance that any one specific person is out today. Four percent sounds negligible.
For five independent jobs each carried by their own single person, that chance does not add up, it multiplies away from one hundred percent: the chance that ALL five people show up is 0.96 to the power five. What is left over — the chance that at least one of the five is missing tonight — is 18.5%. That works out to roughly 55 shifts a year on which at least one of your five critical jobs is not covered by the person who is actually supposed to run it.
It is the same arithmetic an insurer uses to work out the chance of "at least one claim" from a set of small, independent risks: 1 minus (1 minus p) to the power n. What feels like "barely ever" for one job on its own becomes nearly one night in five once you stack five of those risks together.
3. One dark job costs roughly €222 that evening
This site's own guide to staff absence walks through one Saturday night from the 9am phone call to the last guests turned away: one station short-staffed, tickets slipping, dishes quietly dropped from the menu, and eventually 6 covers the host turns away because she will not risk a 45-minute ticket. That is exactly what a dark job means when the one person who covers it is out: not slower — closed, on that one point.
Against the average a guest spends at an independent venue — roughly €37 — that is a lost revenue of roughly €222 for that one evening. Not a scientifically measured figure, but the exact same illustration the staff-absence guide already uses, now put in euros: a real number to attach to a risk that used to just feel "annoying".
And that is before the softer damage the same guide describes: tickets running about 9 minutes longer, dessert and side orders dropping by roughly 15% as the chef quietly simplifies the menu. The 6 covers are the visible part of the bill — not the whole of it.
4. Cross-training one skill takes days; hiring a replacement takes months
Teaching someone a whole new job from scratch takes a full month, according to this site's own onboarding guide: one day to get everything ready, a first week of one station at a time alongside a dedicated buddy, two weeks working solo with short daily check-ins, and a review on day 30. For specialised roles it runs longer still — the guide to staff retention puts full independence for a sous chef or sommelier at 3 to 6 months.
Cross-training is a much smaller problem. You are not teaching a new hire the whole venue, you are teaching an existing team member one extra skill on top of everything they already know about the venue, the team and the guests. The same pattern the onboarding guide uses for a whole job — shadow first, then do it together, then do it alone with feedback — works for one job in a fraction of the time: three shadow shifts, followed by two supervised solo shifts, an illustrative rule of thumb of five shifts of roughly five hours each, focused on that one job rather than a full shift.
Filling a vacancy, by contrast, takes a length of time nothing on this site will pin down for you. This site's own guide to finding hospitality staff is honest about that: good hospitality talent is gone within days to the venue down the street, and a vacancy that "can take months" is a real possibility with no fixed number — it depends too much on role, region and season to put one on it. What does hold: that time comes on top of the 30 days (or 3 to 6 months) of onboarding that still follows, while cross-training starts on day one with staff you already have.
5. That cross-training pays for itself in about 2.1 months
Work through the worked example above. Five shifts of five hours each, at the average gross hourly wage of around €16 this site's staff-absence guide also uses as a rule of thumb, comes out at a one-off investment of €388 to teach a sixth person one of the five single points of failure.
Set against that is what you save. As soon as one of the five jobs has a second person who can cover it, the number of single points of failure drops from five to four — and with it, the chance that at least one falls through tonight, and the number of dark shifts that produces every year. On the numbers in this guide, that is a saving of roughly €2,263 a year, every year, for an investment you only make once.
Divide that investment by the yearly saving and the cross-training pays for itself in roughly 2.1 months — and keeps paying out, year after year, after that. No other move in this guide comes close to that ratio between what it costs and what it returns.
The same numbers as above: one investment, against the annual saving it produces.
The left bar you pay once. The right bar keeps coming back — every year, for as long as that job is covered by only one person. That is the whole reason this ratio is so lopsided.
6. When that one person leaves, the skill simply leaves with them
Staff turnover is already expensive before single points of failure enter the picture: this site's own guide to staff retention in fine dining cites Cornell University research putting the average replacement cost at $5,864 (roughly €5,400) per employee who leaves — recruiting, training and lost productivity combined — and names annual turnover of 30 to 40% in fine dining, against 75 to 150% across hospitality generally.
For an ordinary job, that is the whole story: you pay the replacement cost, and after 30 days (or 3 to 6 months for a specialised role) the new hire is up to speed. For a single point of failure, it is never the whole story. The moment that one person leaves, that specific job does not move from "4% chance of going uncovered tonight" to "a bit higher" — it moves to a hundred percent certain to be uncovered, every night, until someone new is trained. There is no second name to fall back on.
That is exactly why a single point of failure leaving weighs more heavily than the replacement cost alone: the skill was never stored anywhere except in that one person's head, and the knowledge leaves the moment they do. Cross-training in advance is the only thing that prevents that — not by making someone less likely to leave, but by making it far less costly when they do.
7. Three people per job is the point where this risk practically disappears
This site's staff skills matrix tool labels every job by how many people can do it: exactly one name is "single", two names is "thin" (still exposed), and three or more names is "ok". That is not an arbitrary cut-off — the arithmetic behind it shows exactly why.
With one person covering a job (p = 4%), the chance it is completely uncovered on any given night is 4% — roughly 1 in 25 shifts, or about once every five weeks. With two independent people covering it, BOTH have to happen to be out for the job to go fully dark: p × p, or roughly 1 in 625 shifts — already far rarer, but not yet a theoretical zero. With three people — the point at which the skills matrix shows "ok" — that drops to p³, roughly 1 in 15,625 shifts, which on a normal opening schedule is something you would never actually encounter in practice.
That does not mean every job needs three people able to do it — that would make a five-person kitchen impossible. Research on manufacturing process flexibility by Jordan & Graves (Management Science, 1995) shows the same principle at work on production lines: a limited, well-chosen amount of overlap — each worker trained just one job beyond their own, in a chain rather than everyone able to do everything — already captures almost all the benefit of full flexibility. The goal is not "everyone can do everything", it is this modest overlap, placed exactly where it is missing today.
Run your own venue through the calculator
Enter how many single points of failure your team is carrying, how often someone is typically absent, what a dark job costs you, and how many shifts you run a year. The fields already carry the numbers from this guide — overwrite them with your own.
You get the chance that at least one job goes uncovered tonight, how many shifts a year that adds up to, what one dark job costs, and an illustrative yearly cost of that risk if nothing changes.
Single-point-of-failure exposure scan
Five numbers from your own team, and the risk in percent, shifts and euros a year.
—
The ratios in this scan are an illustrative model, not a measurement of your specific venue. Everything runs in your browser; nothing is sent or stored.
Open the staff skills matrix for your own teamTwo numbers worth keeping. The chance at the top is not a guess but the same 1 minus (1 minus p) to the power n insurers run — add another single point of failure and the odds climb faster than most people expect. The yearly cost underneath is not a prediction of what happens this year, but an average across many years — some years it never shows up, some years it lands in one bad week.
What to do with that is not to put three people on every job overnight — a small team cannot afford that. It is to identify the jobs with exactly one name against them, and start with whichever one is most expensive, or most likely, to go dark first.
What to do with this this week, this month and this quarter
Nobody tackles seven numbers at once. This order works, because each step makes the next one measurable.
This week — map your single points of failure
- Open the free staff skills matrix tool and tick off what each team member can do — the tool flags the jobs with exactly one name itself.
- Fill the calculator above in with your own single-point-of-failure count and your own absence rate from the staff-absence guide.
- Rank those jobs from most to least expensive if they went dark tonight — not every job weighs the same.
- Talk the list through briefly with your team: often everyone already knows informally who "the only one" is on each job, it just was never written down anywhere.
This month — train the highest-risk job first
- Pick the most expensive or most exposed single point of failure on your list and schedule three shadow shifts for it with a second team member.
- Follow that with two supervised solo shifts — the same structure as the onboarding guide, just for one skill instead of a whole job.
- Update the skills matrix once those five shifts are done and watch your single-point-of-failure count drop immediately.
- Repeat with the next most expensive job once the first is done — one job a month is a pace most teams can sustain.
This quarter — build the habit, not just the list
- Re-run the scan every quarter: new hires, departures and changed opening hours all shift which jobs are exposed.
- Make cross-training a standard line in every new hire's onboarding plan — it costs almost nothing extra on top of onboarding you are already running.
- Read your staff turnover numbers alongside your matrix: someone who can cover several jobs is often also the person who stays longer.
- Do not aim for "everyone can do everything" — aim for three carriers on every job you have flagged as a single point of failure today, and leave the rest as it is.
The risk was always there — now it has a number
Nothing in this guide asks for a bigger team or a more expensive one. It asks you to ask the same people you already have one explicit question — who can really run this job alone — and then to fix the most expensive of those answers first.
That is exactly the difference between cross-training that happens "eventually" and cross-training that actually happens: without a number, the risk stays a vague feeling that gets pushed back every week. With a number — nearly one night in five, a payback period a little over two months — it becomes a decision that defends itself.
Start with the job that today sits on exactly one name, whose attendance you can predict least. It is not always the most expensive job to lose for a night, but it is almost always the one the next surprise comes from.