Carlos Upgrade | November 8, 2026 | Game theory

Two strangers are told to meet somewhere in New York City tomorrow. They cannot call, text, or agree on a location. How could they possibly find each other?

A map provides thousands of choices. Yet some places and times stand out because each person can imagine the other person choosing them too.

1. The mechanism

Economist Thomas Schelling studied this problem in The Strategy of Conflict (1960). His famous New York meeting example points toward Grand Central Terminal at noon: not because the location is mathematically best, but because it may be especially obvious to people who share a cultural reference.

This is a focal point, often called a Schelling point. When several outcomes could work, people may converge on one because they expect others to notice the same cue.

The logic has two levels. You notice a landmark. More importantly, you believe the other person will notice it, and you believe they will expect you to notice it. The shared expectation can help people coordinate without a message.

That is different from an information cascade. In a cascade, later people can follow earlier people's observed actions. A focal point can operate before anyone sees another person's choice.

In a 1994 experiment by Judith Mehta, Chris Starmer and Robert Sugden, participants made different selections when choosing freely versus when trying to match an unknown partner. The instruction to coordinate changed which options became appealing. That does not make a focal point a guarantee of success.

2. Outside CPG: meeting without a plan

Return to Schelling's New York thought experiment. You might choose the station clock at noon because it is memorable, public, and easy for both people to name. Another pair from a different city, generation or culture might converge on an entirely different landmark.

This is the boundary: salience must be shared, not merely obvious to you. A bright sign that only one person has seen is a poor focal point.

Interests also matter. A 2019 experiment by Stefania Sitzia and Jiwei Zheng found stronger coordination around salient choices when participants' interests aligned. Under substantial conflict of interest, group coordination was not consistently better. A shared cue cannot automatically overcome opposed incentives.

3. What it suggests for beverage packaging

Imagine four fictional lemon drinks in a cooler: a glass bottle, a carton, a plastic bottle, and a pouch. Their fictional brands have different identities, but each uses a recognizable yellow citrus cue.

No industry-wide meeting is required for people to converge on that convention. It can help a shopper and a designer coordinate around an expected meaning without speaking. The analogy is useful, although a shopper choosing a drink is not literally the same as a pure coordination game.

Now imagine launching a citrus product with no recognizable citrus cues. The difference might create distinction. It might also make the flavor harder to find. Neither outcome follows automatically.

My proposed decision would be to test two genuinely different packaging directions under comparable retail conditions: one retains an expected citrus cue inside a distinct brand system; the other deliberately breaks that convention. Measure brand recognition, correct flavor identification, and confusion separately. Include practical compliance and production checks.

4. Challenge the assumption

Assumption: If a design looks more original, it must communicate better.

Not necessarily. Shared visual conventions can reduce decoding work. But the strongest counterargument is real: overusing conventions can produce a wall of generic products. A brand needs recognizable differences too.

The question is not whether to follow every convention. It is which expectation helps the customer, and which one is worth challenging.

5. Your mental operator

Use Find the shared landmark. Before changing a familiar signal, ask: What does our audience already expect? Is that expectation actually shared across our customers? Which part carries useful meaning? What evidence would show that breaking it helps more than it hurts?

6. Reasoning exercise: answer before reading

In an illustrative model, two people independently choose either meeting point A or B. Each picks A with probability 75% and B with probability 25%. What is the probability they select the same meeting point? Compare that with independent 50/50 choices.

Worked answer

They match at A with probability 0.75 × 0.75 = 0.5625, and at B with probability 0.25 × 0.25 = 0.0625. Total: 62.5%. With independent 50/50 choices, matching probability is 50%.

These probabilities are invented for teaching. Real people may be correlated, different groups may notice different cues, and a salient option can fail when incentives conflict. The calculation illustrates why a shared default can help, not how often it does in a particular market.

Keep this: Some conventions are not rules anyone wrote. They are places where separate minds expect to meet.

Research

• Thomas C. Schelling, The Strategy of Conflict (1960): https://books.google.com/books/about/The_strategy_of_conflict.html?id=Bdt-AAAAMAAJ
• Mehta, Starmer & Sugden, “Focal points in pure coordination games: An experimental investigation” (1994): https://research-portal.uea.ac.uk/en/publications/focal-points-in-pure-coordination-games-an-experimental-investiga/
• Sitzia & Zheng, “Group behaviour in tacit coordination games with focal points” (2019): https://research-portal.uea.ac.uk/en/publications/group-behaviour-in-tacit-coordination-games-with-focal-points-an-/

Educational editorial content. Hypothetical examples are labeled in the original edition.