Graveyard, Swing and the Shift Nobody Wants

In a 24-hour market the roster runs day, swing and graveyard, and the last one is filled from whoever is left.

Graveyard gets the newest staff, the thinnest cover, and almost no management attention — because whoever runs the room is asleep during it. Then someone looks at the labour ratio for that block, sees a bad number, and proposes cutting it.

The number is usually bad. It is also usually measured wrong.

Why the ratio misleads

Labour as a percentage of revenue is the standard measure and it breaks on overnight blocks for three reasons.

Fixed cost does not scale down. You need a minimum staffing level to be open at all — someone on the door, someone behind the bar, someone who can call for help. At low volume that floor dominates the ratio, and no amount of efficiency fixes it. The ratio is measuring the floor, not the performance.

Prep and close bleed into it. In a lot of rooms the overnight staff do the deep clean, the restock, and the setup for the day shift. That labour is booked to graveyard and the value shows up on somebody else's shift.

Shift differential inflates the top. If you pay a premium for overnight, the labour number is higher for the same hours worked. That premium is a real cost, but comparing the raw ratio against a day shift that has no differential is comparing two different things.

Correct for those three and the overnight block frequently looks materially different. Sometimes still bad. Frequently not as bad as the raw ratio suggested.

The number that actually decides it

Contribution, not ratio.

Revenue in the block
  minus variable cost of goods
  minus labour that would not exist if you closed
  = contribution

The key phrase is "would not exist if you closed." If the graveyard crew does the restock, closing the block does not remove that work — it moves it to a shift with different staffing and possibly different overtime exposure.

Rooms that run this calculation properly sometimes find the overnight block is contributing modestly and doing all the prep. Closing it produces a saving on paper and a cost in practice.

What graveyard is actually for

Three different things, and rooms confuse them.

Trade. There are guests and they spend. Measure it normally.

Presence. Being the room that is open when others are not builds something that shows up on other shifts. Hard to measure and real — but it is a marketing claim, so require some evidence rather than accepting it as given.

Production. The work that has to happen and is cheaper overnight. Genuine value, frequently booked to the wrong shift.

Write down which of the three you are running. A block run for production should be staffed and measured completely differently from one run for trade, and rooms that have not decided end up doing both badly.

The staffing problem underneath

Graveyard is filled last, so it gets the least experienced staff, so it runs less well, so the numbers look worse, so it gets less attention.

That loop is worth breaking deliberately, because it is not self-correcting.

A pay differential helps but does not fix it. What moves it more reliably is rotating some experienced staff through, and giving the block a named owner who is actually awake during it — someone whose job includes the overnight, not a day manager who reads a report.

⚠️ Watch the incident rate. Overnight blocks with the thinnest cover and the least experienced staff carry the most risk, and incident cost does not show up in the labour ratio at all.

Before you cut it

Three checks, in order.

One — recalculate the contribution properly. Strip out the labour that survives closure, and account for the differential separately.

Two — find out what production moves. List the tasks currently done overnight and price them on the shift they would move to. Include overtime exposure.

Three — test it before committing. Close it on the two slowest nights for four weeks and watch what happens to the shifts either side, not just to the block you cut.

Most cuts that fail, fail on check two — the work was invisible until it landed somewhere else.

The thing that shows up two months later

When an overnight block closes, the effect on the day and swing shifts appears with a delay.

The prep that used to be done arrives on a shift that was already at capacity. Service slows slightly. Nobody attributes it to a decision made two months ago, and the room concludes it has a staffing problem on a completely different shift.

Write down what you expect to happen before you cut, and check it at four and eight weeks. That single note is the difference between a decision you can learn from and one you cannot.

Three numbers to hold

Contribution by shift block, with labour split into "survives closure" and "does not."

Task list by block — what production actually happens overnight, priced on its alternative shift.

Incident rate by block. The cost that never appears in the labour ratio and that gets worse exactly where cover is thinnest.

Where the record has to sit

Shift blocks, sales, staffing levels, and incidents need to sit against the same night, or the contribution calculation is a spreadsheet exercise somebody does once and never repeats.

tasteck keeps shift assignments, sales by hour, and incident records on the same night's record, so contribution by block is readable rather than reconstructed.

Graveyard is either earning, producing, or neither. It is worth knowing which before it is cut on a ratio that was never measuring the right thing.