Two bartenders serving and nobody restocking. Why the support role is the first one cut and the first thing that breaks service, how to see it in your own numbers, and what a bar back is actually worth per hour.

Saturday. Two bartenders on, both good, both fast. By midnight there is a queue three deep and neither of them has poured a drink in ninety seconds, because one is in the cellar and the other is emptying a bin.

Nobody made a decision to have this happen. The bar back shift was cut three weeks ago to save money on a night that turned out busy, and the cut has never been revisited.

Why the support role is always the first cut

It is the only role on the rota whose absence is invisible in advance.

Cut a bartender  — the gap is obvious before the shift
Cut a door       — the gap is obvious at the door
Cut a bar back   — nothing looks different until 23:30

⚠️ The bar back is the only cut that does not show up until it is too late to reverse.

And it always looks like the cheapest one

The support role is usually the lowest-paid on the rota, so cutting it looks like the smallest saving with the smallest risk. The reverse is true: it is the cheapest hour on the rota and the one that protects the most expensive ones.

⇒ The Shift You Cancelled And What It Cost

What the role actually is

Not "helping." Six specific things, each of which otherwise takes a bartender off the bar.

1  Keeping the speed rail and fridges stocked
2  Ice
3  Glassware in and clean glassware out
4  Bins and spillage
5  Cellar runs and gas changes
6  Fetching what has run out

⭕ Every one of those, done by a bartender, costs a queue.

The one that costs the most

Glassware. When clean glasses run out, the bar stops completely, and it stops at the busiest moment because that is when it happens.

⇒ Most venues discover their glassware capacity is the real limit only on the night it fails.

The one nobody counts

Ice. It is heavy, it is far away, and it runs out silently. A bartender doing an ice run is off the bar for four to six minutes at the peak.

Seeing it in your own numbers

You do not need to observe a shift to find this. It shows up in the takings shape.

The dip that is not a demand dip

Pull hourly sales for a busy night and look for a flat or falling hour in the middle of the peak.

21:00  rising
22:00  rising
23:00  flat        ← demand did not fall here
00:00  rising again

⚠️ Demand does not dip in the middle of a peak. A flat hour inside a rising night is nearly always a service-capacity ceiling, not a demand ceiling.

⇒ Your Quietest Hour

The comparison that settles it

Take two similar nights, one with support scheduled and one without, and compare the peak hour.

With support     — peak hour keeps rising
Without support  — peak hour flattens

⭕ Two nights is not proof, but it is enough to justify testing it properly.

⚠️ Be careful that the two nights are actually comparable — same night of week, no event, similar weather. Otherwise you are measuring something else.

⇒ What The Weather Does To Your Week

What the hour is worth

The arithmetic is simple and most owners have never done it.

Bartender serves roughly N drinks per hour at full flow
Each restocking interruption removes 4-6 minutes
Six interruptions across a peak = roughly half an hour of one bartender
Half an hour at peak, at your average spend

⇒ Compare that to one support hour at your lowest hourly rate.

⚠️ In most venues the support hour pays for itself several times over on the peak night, and does not on the quiet ones. That is the actual answer: it is a peak-night role, not an every-night role.

Which is why the blanket cut is wrong in both directions

Cut on every night  — peak nights lose more than the saving
Kept on every night — quiet nights carry a cost that earns nothing

⭕ Schedule it to the shape of the week, not to a policy.

⇒ Graveyard, Swing, And The Shift Nobody Wants

The second thing the role does

Beyond service speed, the support shift is the only entry point most bars have.

It is where you find out whether someone is reliable
It is where someone learns the room before learning the drinks
It is the cheapest way to test a hire

⚠️ A venue with no support shifts has no way in. Every new bartender is then hired cold, from an interview, which is a far worse predictor.

⇒ Three Weeks Into A New Hire

And it is the succession plan

The person who has spent three months keeping the bar stocked already knows where everything is, how the night flows, and who the regulars are. Promoting them costs nothing and risks nothing.

⇒ Your Busiest Person Is A Single Point Of Failure

What to do this week

One: find the flat hour

Pull hourly takings for your last four busy nights. Look for the flat spot inside the peak. Note the hour.

Two: watch that hour once

Stand near the bar for that one hour and count how many times a bartender leaves the bar, and for how long.

⚠️ Count it. Estimating this is unreliable in both directions.

Three: schedule support to cover that hour, for two weeks

Not the whole shift. The hour before and the hour of the flat spot.

Four: compare the same hour

Flat spot gone      — you found it
Flat spot unchanged — the ceiling is somewhere else (terminals, glassware capacity, space)

⇒ When The Card Machine Is The Queue

What this looks like when it is working

The bar test

At peak, are both bartenders behind the bar?

Working     — yes, continuously
Not working — one of them is somewhere else every few minutes

The glass test

Does the bar ever run out of clean glasses?

⇒ If yes, that is either a support gap or a glasswasher capacity limit. Both are cheap to find and neither is a staffing attitude problem.

The rota test

Look at the rota. Is support scheduled by night shape, or by habit?

The same shape in other roles

The support role is the clearest case, but the pattern is not confined to the bar.

The glass collector

Same economics, same invisibility. When glasses are not collected, the bar runs out of clean ones and the bar stops — which reads as a bar problem, not a floor problem.

The second door

Cut, and the queue at the entrance grows. The cost lands as people who walked away, which is the least visible loss a venue has.

⇒ The Door Count That Tells You Nothing

The person who runs the cellar

⚠️ Every venue has one person who knows where things are. When that person is not on, everyone else spends the night looking.

⇒ That is not a support-role problem. It is a single-point-of-failure problem wearing the same coat.

The three-hour version

You rarely need the role for a whole shift. You need it for the window where the ceiling appears.

Working out the window

Hourly takings for four busy nights
The hour where the rise flattens
That hour, plus the one before it

⭕ Two or three hours, not eight. That is a materially different cost.

Which hour to start

⚠️ Starting at the flat hour is too late — the bar is already behind by then.

⭕ Start ninety minutes before it, so the stocking is done before the peak rather than during it.

⇒ The Run Of Show For A Busy Night

The end of the shift matters too

The same person, kept for thirty minutes after close, takes a meaningful chunk out of the cash-up and the clean-down.

⇒ The Cash-Up That Takes An Hour

What the role needs to be given

A support shift with no equipment and no information is just a person carrying things.

The information

Where everything is kept
What the par levels are
What to do when something runs out and there is no more

⇒ Par Levels And The Money Sitting In Your Store

The authority

⚠️ If they have to ask a bartender before every action, they are interrupting the person they exist to protect.

⭕ A short list of things they may do without asking. That is the whole design.

The route

Cellar to bar, with a door somebody has to open — the role is half as effective

⇒ Physical routes matter more for this role than any other, and they are usually nobody's problem.

Where venues get the cost wrong

Counting the wage and not the peak

The hour is cheap. The bartender minutes it protects are the most expensive minutes in the week.

Assuming it scales with headcount

Two bartenders — the interruptions are shared
Three bartenders — the interruptions triple, the support need does not

⚠️ The support requirement scales with throughput, not with staffing.

Cutting it on the wrong night