Ask three restaurants what they tip out to the bar and you will get three numbers, all delivered with total confidence. Ask where the number came from and the answers thin out considerably: it was inherited from a previous manager, or copied from the place down the road, or picked because it sounded about right.
That is survivable. What is not survivable is the part nobody checks — whether the tip-out is a share of sales or a share of tips. Those two look almost identical in a policy document and behave completely differently on a slow Tuesday.
The base is the decision
A tip-out is a transfer from the people the guest handed money to, towards the people who made the service possible but were never at the table. Runners, bartenders, hosts, kitchen. The mechanism has two parts, and only one of them gets argued about.
The percentage is what teams debate. The base — the thing the percentage is taken from — is what determines whether the arrangement holds up when trade is poor.
| Base | Moves with tips | Can exceed tips | Survives a bad night |
|---|---|---|---|
| A share of the tip pool | Yes | No | Yes |
| A share of sales | No | Yes | No |
A sales-based tip-out also needs reliable point-of-sale data, where a share of the pool needs only the pool. But everything that matters follows from one sentence: it levies a fixed obligation against a variable income.
What a sales-based tip-out does on a bad night
Take a server who closes €1,800 of sales and owes 3% of it to the support pool. That is €54, and on a good night it is unremarkable.
3% of sales on a good night
- Sales closed
- €1,800.00
- Tips received
- €190.00
- Tip-out owed (3% of sales)
- −€54.00
- Server keeps (72% of the tips)
- €136.00
Now the same section on a night when a large table pays by card and rounds to the nearest euro, and a second one leaves nothing at all. The sales barely move. The tips collapse.
The same 3% of sales on a bad night
- Sales closed
- €1,700.00
- Tips received
- €62.00
- Tip-out owed (3% of sales)
- −€51.00
- Server keeps (18% of the tips)
- €11.00
The percentage did not change. The share of the server’s actual earnings went from 28% to 82%. Push the sales a little higher and the tips a little lower — a large party that tips nothing is enough — and the tip-out goes above 100%, at which point the server is paying out of pocket for having worked the busiest section.
A percentage of the tip pool has none of this behaviour. If the pool halves, every share halves. Nobody is happy, but nobody is being charged a fee to have shown up.
Deriving the percentage instead of inheriting it
Once the base is the pool, the percentage becomes an arithmetic question rather than a matter of taste. A tip-out is a shortcut for a points split, so the way to size it is to run the points split you would have run and read the answer off.
Take a typical evening: three servers on eight hours, one runner on six, one bartender on six.
role share = role points ÷ total points
Weight each role's hours by its point value, then divide that role's points by the total. The result is the percentage a tip-out would have to be to match a points split.
What a points split gives the support roles
- Servers, 3 × 8 hours × 1.0
- 24.0 points
- Bartender, 6 hours × 1.0
- 6.0 points
- Runner, 6 hours × 0.7
- 4.2 points
- Total points
- 34.2
- Bartender share (6.0 ÷ 34.2)
- 17.5%
- Runner share (4.2 ÷ 34.2)
- 12.3%
- Tip-out that reproduces the split
- 18% bar · 12% runner
Round to whole percentages and you have a policy: eighteen to the bar, twelve to the runner, the remainder split between the servers by hours. It is defensible in a way an inherited number never is, because you can show anyone the shift it was derived from.
One number per role, not one number total
“We tip out 25%” is a policy that hides a decision. The bar and the runner did different amounts of work for the same table, and lumping them together means the split between them gets settled informally — which is to say, by whoever is more assertive.
Setting a number per role costs nothing extra, since the derivation already produces one. It also makes the review tractable. When you hire a second runner, only the runner line needs revisiting.
The statutory code of practice that accompanies the UK tipping legislation lists the factors an employer should weigh when deciding how tips are shared: the type of role, hours worked during the period the tips were paid, individual and team performance, seniority, customer intentions and length of service[1]. It is not binding outside the UK, and it prescribes no percentages at all. What it is useful for is as a checklist of the things a defensible share is allowed to depend on — a per-role number can point at several of them, while a single lump can point at none.
The kitchen is a different question
Sharing with the kitchen is the one case where the derivation gets awkward, because a kitchen point value is not really a measure of anything. Nobody can say whether an hour of plating is 0.5 or 0.7 of an hour of service, and the argument for a specific figure is always circular.
What helps is deciding the principle before the number:
- If the food is a large part of why guests tip, the kitchen has a claim on the pool and the number should be big enough to be felt — a token 5% reads as an insult rather than a share.
- If the kitchen is already paid a premium precisely because it does not receive tips, adding a share on top is a wage decision dressed up as a tip policy, and it belongs in a wage conversation.
Teams that do share tend to land between a fifth and a third of the pool. Whichever end you pick, tie it to a review trigger rather than a date: a menu change, or a change in kitchen headcount. A share set when the kitchen was two people rarely still fits when it is five.
Where the ceiling sits
There is no legal cap on a tip-out, but there is a practical one. Once more than roughly a third of the pool is leaving the section that earned it, the arrangement has stopped being a tip-out. It is a pooled split with extra steps, and it will keep generating friction precisely because it is described as something else.
That is not an argument against pooling. It is an argument for calling it what it is: if two-fifths of the money is going to support roles, run a pool and split it by points, where every share is visible and nobody has to think of themselves as paying somebody else. The arithmetic can be identical. The way it reads to the people it applies to is not.