Covers, Friday dinner
58 → 68
Same room, same service hours — layout C
The floor was set at fit-out, five years ago, for a town that booked in fours. Today 58% of parties are twos — and every two seated at a four-top strands two chairs the book counts as sold. The optimiser re-runs the room against the bookings actually walking in the door, and prices what a refit is worth before a dollar is spent. Read by Eleanor Marlowe, Owner
Every restaurant runs this optimisation exactly once — at fit-out, on instinct, before a single booking exists. Then the booking mix drifts for years and nobody re-runs the sum, because re-running it means moving furniture.
The house already holds the data to do it properly: a year of party sizes from the phone and chat agents, spend per head from the till, and the room's measured geometry. The optimiser searches thousands of legal layouts — legal meaning every fire-egress aisle, accessibility route and service path is respected — and scores each one in covers, dollars per head and dollars per square metre.
It doesn't redesign the room's character. It prices the gap between the room you have and the room your bookings are asking for — and hands you the budget that gap justifies.
The finding: the most covers isn't the most money. Layout B seats the most people; layout C — fewer, dearer seats, with the chef's counter selling the tasting and the pairing — wins on both revenue and dollars per head. Maximise covers per dollar of guest, not covers alone.
The uplift is the budget's ceiling, worked backwards: at a 65% gross margin the layout returns ~$188k of contribution a year, so a $68k fit-out clears in 4.3 months. Anything under ~$94k pays back inside six.
"Layout C returns $1,112 more per service. At 260 services a year that's $289k gross, ~$188k contribution — a $68k refit pays back in 4.3 months. Spend up to $94k and you're still inside a six-month payback."
The operator isn't handed a fait accompli. Nudge the banquette, keep the window deuce Eleanor loves, drop the counter to four seats — every drag recomputes covers, $ per head and the payback live. The optimiser proposes; the owner decides what the room is.
Pushed for 60 seats, an unconstrained packer will happily oblige — and hand you a room that fails its fire inspection and exhausts the floor staff. This one treats the codes as hard walls, not suggestions. 60 seats is not a layout this room can legally hold; the guard capped the search at 54 and explained every centimetre it refused to give.
Fire egress, 1.2 m main aisle — SCDF Fire Code escape-route width to both exits. The 60-seat packing narrowed it to 0.9 m. Non-negotiable; the optimiser will not trade escape width for chairs.
Accessible route, 900 mm — BCA Accessibility Code path from door to accessible seating and the WC. Two proposed deuces pinched it. Removed before the layout was ever shown.
Service path to the pass — 750 mm runner's line from kitchen to the banquette, a house rule rather than a code. Tighter than ideal at one pinch; flagged for the walk-through, kept with a warning.
The rule: code clearances are constraints, never variables. The optimiser searches only the layouts that pass SCDF egress, BCA accessibility and the house service paths — then maximises within them. A room that seats 60 and fails inspection is worth $0 a head.
This page is a concept. The room, layouts and dollars are hand-built to tell one refit's story honestly; nothing here ran against a measured floor. The shape of the real thing: a constraint solver over the room geometry — table footprints, SCDF egress aisles, BCA accessibility routes, service paths — scoring candidate layouts against the party-size distribution the booking agents (UC1 / UC2) already collect and the spend-per-head the till already knows. The language model narrates trade-offs and writes the capital case; the optimiser does the arithmetic. It's the rare use case that monetises data the house already has at a moment — fit-out, refurb, new site — when real money is about to be spent on instinct.