The fifteen-minute window

Concession demand at a cinema is not a smooth curve. It arrives in pulses timed to showtimes. Most customers buy between roughly ten minutes before the advertised start and the end of the trailers. For a session with 200 seats and a 40 percent purchase rate, that is 80 transactions crammed into a window that, in practice, is about fifteen minutes long. If two screens start within ten minutes of each other, the pulses overlap.

This is why stagger matters. A multiplex that starts all its evening shows at 7:30 has built itself a single fifteen-minute traffic jam. One that staggers starts by ten or fifteen minutes has spread the same demand across an hour. No amount of extra staff fixes the first cinema as cheaply as a schedule change does.

Little's Law and the length of the line

The single most useful result in queueing theory is Little's Law: L = λW. The average number of people in a system equals the average arrival rate multiplied by the average time each spends there. It holds regardless of arrival pattern or service distribution, which is why operations textbooks such as Cachon and Terwiesch's Matching Supply with Demand open with it.

Apply it to the stand. If customers arrive at 6 per minute and each spends 4 minutes from joining the line to walking away with a tray, there are on average 24 people in the lobby queue. If you can only cut that to 3 minutes of wait plus service, the line drops to 18. Little's Law does not tell you how to cut the wait, but it tells you exactly what a given wait costs in crowd size, and crowd size is what customers see when they decide whether to join.

The subtler point is that W includes service time. A cashier who takes 60 seconds per order instead of 45 has not slowed the line by 25 percent; because arrivals keep coming during the extra seconds, the queue grows nonlinearly as utilization approaches 100 percent. In the standard M/M/c model the expected wait explodes as the ratio of arrival rate to total service capacity approaches one. A stand running at 95 percent utilization during the rush is not "efficient"; it is one sneeze away from a lobby full of people who give up.

Multi-bottleneck flow: it is never just the cashier

Real stands are not a single queue. An order flows through several stations: the till, the popcorn warmer, the soda fountain, the hot-food station for nachos or hot dogs. Each has its own capacity. The throughput of the whole stand is the throughput of its slowest station for the current mix of orders, a result operations managers know from Goldratt's theory of constraints.

That has an unintuitive consequence: the bottleneck moves with the menu. A crowd buying popcorn and soda is limited by the popper's output and the number of dispensers. A crowd buying combos with nachos is limited by the cheese warmer. Adding a second cashier to a stand whose popper is the constraint adds nothing but a second person waiting on the popper.

The model has to track each resource separately and let the slowest one bind. Any simulation that reduces the stand to "staff × speed" will systematically overstate throughput during the exact minutes that matter.

Abandonment: the sale you never see

Customers who look at a long line and go straight to their seats are the invisible cost of under-capacity. Queueing researchers call it balking (refusing to join) and reneging (joining, then leaving). Both rise steeply with visible queue length, and both are especially severe at a cinema because the customer has a hard deadline: the film starts whether or not they have their popcorn.

The financial damage is amplified by the margin structure. Each lost concession sale is a lost dollar of nearly pure profit, not a lost dollar of revenue. A stand that turns away 15 percent of its demand at peak has thrown away a far larger share of its weekly profit than the raw number suggests.

Staffing the peak without paying for the trough

The queueing results above push toward more capacity, but capacity costs money every hour it exists, and the rush lasts fifteen minutes out of every ninety. A stand staffed for the peak is idle most of the evening. This is the classic service-industry trade-off: labor is a fixed cost within the shift, demand is a pulse.

Operators manage it three ways. First, cross-training: a staff member who can run the till, pop corn and pull drinks can move to whichever station is binding, which effectively raises the capacity of the bottleneck without adding a body. Second, pre-production: popping ahead of the rush so the popper's output rate stops being the constraint during the window, at the cost of freshness and spoilage. Third, self-service: fountain drinks and pre-packaged items that remove a station from the flow altogether.

Each has a cost, and the choice between them depends on the menu mix and the showtime pattern. That interaction is a decision, not a formula, and it is one of the more satisfying parts of running a stand once you can see the numbers.

Inventory that spoils

The other half of the concession problem is stock. Popcorn kernels and syrup keep for months, but popped corn, hot-held food and some fresh items have a shelf life measured in hours or days. Over-order and you throw margin in the bin; under-order and you hit the abandonment problem above.

The right mental model is batches with expiry. Every purchase creates a lot with a quantity and a use-by date. Demand consumes the oldest lot first. Whatever is left when a lot expires is a write-off. Because demand is pulsed and uncertain, the optimal order quantity is a newsvendor problem: balance the cost of overage (spoilage) against the cost of underage (lost high-margin sales). With margins this high, the newsvendor answer usually says to err on the side of too much, up to the point where storage capacity or cash runs out.

The National Association of Concessionaires publishes benchmarks and best practices on this exact trade-off, and the consistent lesson is that stock-outs during the rush are more expensive than the spoilage most operators fear.

Combos and the arithmetic of the upsell

Combo pricing is often presented as a marketing trick. It is better understood as a throughput tool. A combo collapses three decisions into one, which shortens the transaction; it standardizes the order, which lets staff pre-assemble; and it raises the average ticket, which lifts per-cap even when the discount is real. A stand that sells combos to half its customers processes the rush faster and earns more from it.

The arithmetic is unforgiving in the other direction too. Every combo that includes a slow item, hot food that must be prepared to order, for instance, drags the bottleneck back to the slowest station. The best combos are built from the fastest items with the highest margins, which is why the industry's default bundle is popcorn and a fountain drink and not nachos and a coffee.

Simulating the rush

A convincing concession model therefore needs, at minimum:

• Pulsed demand keyed to each session's start time and audience, not a flat hourly rate.
• Per-station capacity for tills, poppers, dispensers and hot-food equipment, with the slowest binding for the current order mix.
• Abandonment that rises with queue length and time-to-showtime.
• Batch inventory with arrival dates and expiry, consumed oldest-first.
• Storage limits so that "just buy more" is a real constraint, not a free option.

Get those right and the interesting strategic decisions fall out on their own. Staggered showtimes become valuable. A second popper can be worth more than a second cashier. And a cheap ticket that fills the room with hungry customers can be the most profitable pricing decision of the week, provided the lobby can serve them.