System Intelligence vs. Everyone About How It Works Apply for Access →
← Intelligence Feed
Origin Story

What 80,000 Concurrent Users Taught Us About Your 80-Seat Restaurant

The consumer behavioral knowledge from venues far larger than a restaurant that no restaurant platform has ever seen

March 5, 2026 6 min read superGM Intelligence Team
behavioral datastadium operationscrowd sciencecorpus

The first time we ran operational analytics for a 68,000-seat venue, the problem we were asked to solve was concession revenue per cap. The venue was significantly below peer benchmarks. Standard analysis pointed at pricing, product mix, and queue management. These were real factors.

The factor nobody had modeled was time.

The Time Problem at Scale

In a stadium, the majority of concession revenue is generated in four discrete windows: pre-event, first intermission or half, second intermission or half, and post-event. Each window is between eight and twenty-two minutes long. Outside these windows, concession velocity drops to near zero.

The implication is stark: a concession operation at a stadium has approximately sixty to ninety minutes of viable selling time across a three-to-four hour event. Every minute of inefficiency in that window is a direct revenue loss that cannot be recovered. There is no "well, we'll catch it later." Later doesn't exist.

This time pressure — combined with the volume, the crowd dynamics, and the operational complexity of running forty concurrent service points simultaneously — produces a laboratory for operational intelligence that has no equivalent in the restaurant sector.

What Scale Teaches

At up to 80,000 people at once, individual human behavior becomes statistically predictable in ways that are invisible at the scale of a 150-seat restaurant. You can observe, with high confidence, what a crowd will do when a specific type of event occurs. You can model the spending response to emotional triggers — a surprising outcome, a prolonged delay, a moment of exceptional quality — with enough precision to deploy resources ahead of the response rather than in reaction to it.

You learn that price sensitivity is not a fixed property of a product. It is a function of emotional state, crowd density, perceived scarcity, time pressure, and social proof — and all of these factors can be read in real time from behavioral signals if you know what to look for.

You learn that crowd energy propagates through physical space in predictable patterns. You learn the difference between a crowd that is engaged and one that is losing engagement — not from survey data, but from observable behavioral signatures that appear in spending patterns, movement patterns, and communication patterns before they appear in any satisfaction metric.

The Translation to Restaurants

A restaurant dining room is a crowd. It is smaller, more intimate, and slower-moving than a stadium crowd. The emotional dynamics are less intense. But the underlying mechanics — social contagion, price sensitivity modulation by context, behavioral signatures of engagement and disengagement — are the same.

The corpus we built across fifteen years of venues far larger than a restaurant contains patterns that are directly applicable to restaurant contexts. The guest who has been on their phone for eighteen minutes without ordering another drink is exhibiting a departure signature we have seen across trillions of data elements in other contexts in other contexts. The table that is visibly elevated — laughing, gesturing, ordering with enthusiasm — is producing a social energy signal that adjacent tables will respond to in a predictable time window.

The restaurant platforms built from restaurant data alone will never see these patterns. Not because they lack capability. Because the patterns don't exist in restaurant data.

They exist in ours.

Related Intelligence

MORE TO READ.

Application Review

MOST OPERATORS
WHO APPLY
WILL NOT BE SELECTED.

We work with operators whose operation, culture, and competitive position fit what we built this for. We review every application individually. We select from the backlog.

If you are reading this because a competitor sent it to you, they may already be in production. We don’t confirm or deny active deployments.

Applications reviewed individually · Not all are accepted