How to build a team that does not leave every four months

Put a number on turnover first

Most restaurant owners describe staff turnover as frustrating. Very few describe it as expensive, because very few have costed it.

Do that first, because the number changes how everything else on this page reads.

A worked example. An experienced captain on ₹22,000 a month leaves.

Roughly one and a half months of that person's salary, and none of it appears as a line item anywhere.

Now assume four such departures a year. That is ₹1.2 lakh, invisible, recurring — against a restaurant whose net margin might be ₹1.5 lakh a month. Turnover is not an HR annoyance. It is a cost line hiding inside several other cost lines.

Once that is on paper, a ₹1,500 monthly raise to keep someone, or an afternoon spent building a proper first-week plan, stops looking like an expense and starts looking like arithmetic.

Why people actually leave

Ask someone why they are leaving and they will usually say pay. Pay is real, and it is also the socially easy answer.

The recurring structural causes, in roughly the order they show up:

Rosters published late or changed at short notice. This is the big one and it is almost free to fix. A person who finds out on Saturday evening what they are working on Monday cannot plan a life. They will leave for a job paying the same amount with a predictable schedule.

No visible path. A commis who cannot say what the next role is, what it pays, or what would earn it has no reason to stay past the first better offer. The path does not have to be fast. It has to be stated.

Unpredictable management. Not harshness — unpredictability. A manager whose reaction to the same mistake differs by day teaches the team to hide problems, which is worse operationally than the mistakes were.

Favouritism in shifts. Who gets the good sections and the weekend shifts is watched far more closely than owners realise, and perceived unfairness here poisons a team faster than pay differences do.

Being set up to fail on day one. Covered below. It is the most fixable of all of these.

Hiring: the parts that matter

The full treatment — where candidates actually come from, what a trial shift should test, and the first week that decides whether they stay — is in how to hire restaurant staff who actually stay. Three principles worth stating here:

Hire before you are desperate. Panic hiring on the day someone quits is how you end up filling the same role three times a year. If your kitchen runs on the assumption that nobody will ever leave, it is running on a bad assumption.

Trial shifts should test the actual job. Not a conversation about the job. Put them on the floor or the line for a paid trial during a real service and watch how they behave when something goes wrong — because something will.

Check the thing you care about. If reliability matters most, ask a previous employer about attendance rather than about skill. Skill can be trained in weeks. Turning up cannot.

Roles: write them down

Most restaurant teams operate on an oral tradition of who does what. It works until someone is absent, a new joiner arrives, or two people each assume the other did the thing.

Write one page per role. Not a legal document — a working one:

• What this person owns during service
• What they own before and after service
• Who they escalate to
• What "done well" looks like, specifically

The value is not the document. It is that writing it forces you to notice the tasks currently owned by nobody, and the ones owned by two people who each assume the other does it. The closing tasks are usually the worst offenders, which is why opening and closing checklists work best when every line has a named role beside it rather than being addressed to "the team".

The first seven days

This is where retention is won or lost, and it is where most restaurants perform worst — because the first week falls during service, when nobody has time.

A workable structure:

Two rules that matter more than the schedule:

Name the trainer. "The team will show them" means nobody will. One named person, who knows it is their job, and whose own load is adjusted for that week.

Do tool and menu training off the floor. A new captain learning your ordering system during a rush will revert to paper at the first confusion — and then you are running two systems and the training never completes. One hour, real devices, real menu, before a live service. This applies whether the tool is a notepad convention or software; the point is that learning under pressure does not stick.

Rostering without the weekly argument

Roster conflict is rarely about who works when. It is about people finding out too late.

Staff to demand, not to availability. Pull orders by hour of day over four weeks, split weekday and weekend. Most restaurants discover their peak is 45–90 minutes later than the roster assumes, and that they are overstaffed early and thin late. Moving the same hours costs nothing. The method is in restaurant analytics.

Publish on a fixed day, every week. Wednesday for the following week, say. The specific day does not matter. The consistency does. This is the cheapest retention improvement available to most restaurants and it costs nothing but discipline.

Set a request deadline before that day and hold it. Requests before the deadline get accommodated where possible; requests after it get accommodated only if someone swaps. Applied evenly, this ends most of the argument.

Track who gets the good shifts. Weekends and prime sections are watched. If the distribution is not roughly even over a month, someone has noticed even if nobody has said so.

Common mistakes

Hiring in a panic. Produces the churn it was meant to solve.

Training during a rush. The worst possible conditions for learning, and it teaches people that asking questions is an imposition.

"The team will show them." Unowned onboarding is no onboarding.

Treating pay as the only lever. A raise buys a few months. Predictable rosters and a stated path buy years, and cost less.

Promoting your best cook to manage. Cooking well and running people are different skills. Promote deliberately, with training, or create a senior technical role instead.

No exit conversation. Fifteen minutes with someone who is already leaving and has no reason to be diplomatic is the most honest feedback you will ever get about your restaurant.

What to do this month

1. Cost your last departure using the table above. Put a real number on it.
2. Write one page per role. Owned tasks, escalation, what good looks like.
3. Pick a roster publication day and hold it for eight weeks without exception.
4. Build a seven-day onboarding plan with a named trainer and tool training off the floor.
5. Check shift distribution over the last month. Is it as even as you think?
6. Do an exit conversation with the next person who leaves, and actually write down what they say.

The short version

• Replacing an experienced staff member costs roughly one to two months of their salary, hidden across several other cost lines.
• People leave over unpredictable rosters, no visible path and inconsistent management at least as often as over pay.
• Write roles down. It exposes the tasks currently owned by nobody or by two people.
• The first seven days decide retention. Name a trainer and do tool training before service, not during it.
• Publish the roster on the same day every week. Cheapest retention fix available.
• Roster to your actual demand curve, not to a belief about when the rush lands.