Kitchen throttle: the pacing rule that saved our Fridays

An 80-cover room can do 80 covers at 7pm and 80 at 9:15 — but not 80 at 8pm without collapsing tickets. Pacing isn't about tables; it's about how many mains hit the pass per 15-minute window. Ask any head chef who's worked a Saturday at Hawksmoor, or read Sean Brock's Heritage kitchen notes — the ceiling is always the pass, not the dining room.

We measured ours empirically: 14 mains per 15 minutes is smooth, 18 is the ceiling, 22 kills the section. So we set the availability engine to allow max 12 new covers seated per 15-min slot, staggered by section. The number is trivial to arrive at once you have a service logging tool — count the tickets that go out of the pass in 15-minute windows across four services, note the point at which the fry section starts calling for expo, and back off by 15%.

The insight most reservations systems miss is that 'table available' and 'kitchen available' are two different questions. OpenTable's engine, ResDiary's engine, SevenRooms' engine all default to answering the first question. All three let you configure the second, but the config is buried three menus deep. Nine out of ten operators we've onboarded from these systems had never touched it. Their Fridays looked like their Fridays for a reason.

Sunday lunch is one seating, 90-minute turn, kitchen glides. Saturday dinner is three waves, 105-minute turn, kitchen sprints. Same room, different physics. Configure them separately — every time.

The Sunday lunch service at St. John in Smithfield is a masterclass in this: they'll take 100 covers across a single seating window because the menu is simplified, the sides are par-cooked at 12:30, and the pass isn't ticket-driven — it's tray-driven. That's a completely different throttle profile from their weekday dinner. Same kitchen, same team, entirely different pacing engine.

The pattern to steal: for each service, define (a) the seating window (open → last order), (b) the max covers per 15-min slot, (c) the target turn time, and (d) the 'do not exceed' concurrent covers at any given minute. Those four numbers, per service, are the whole model. Everything else your reservations system offers on top is polish.