Theory-of-the-Game — Critical Mass Thesis Building
Build a specific model for the path to critical mass before smart critics see it
The Theory-of-the-Game is the specific mental model a founder must have before launching a network-effects business into a market where smart people believe it will fail. It is the intellectual core of Hoffman's contrarian idea selection framework — without it, contrarian pursuit is just stubbornness, not strategy.
The theory must explain a non-obvious path from thin early traction to critical mass: who the early explorer cohort is (the 10–100 of every 1,000 users who immediately see the potential), what causes them to self-identify and stay, how their behaviour drives the network's growth even while the majority of potential users remain skeptical, and what the product looks like when critical mass is reached and value becomes self-evident to the skeptical majority.
The key insight is that the theory enables persistence through the exploration phase — the period when the product has some early traction but the majority of users behave exactly like the smart critics predicted (they sign up and then don't use it, or don't sign up at all). Without a pre-existing theory that explains why early non-use is expected and what comes after it, a rational founder looking at the data will abandon the product before the explorer cohort has had time to drive network growth.
- A contrarian idea without a theory of the game is stubbornness, not strategy — the theory is what distinguishes a bet from a guess.
- The explorer cohort (10–100 of 1,000 early users) drives the network toward critical mass; majority non-use during this phase is expected, not failure.
- The theory must predict why critics are wrong about the trajectory (path to critical mass) even when they are right about the current behaviour (most users don't engage).
- Critical mass transforms a product's value proposition from invisible to self-evident — the theory must specify what critical mass looks like and how it will be reached.
- Persistence through the exploration phase is only rational if you have a pre-existing theory that explains early data patterns as expected rather than surprising.
- Identify the explorer cohortBefore launching, model who the 10–100 of 1,000 early users are who will immediately see the network's potential. What drives them? Why are they in the explorer cohort when the majority are not? What do they look like so you can find and prioritise them?Pro tipExplorer cohorts are often domain experts, early adopters in adjacent fields, or people who have previously experienced a comparable network in a different context. They see the endgame because they have pattern-matched from prior experience.
- Define critical mass preciselySpecify the threshold at which the value proposition becomes self-evident to the skeptical majority. This is market-specific: for payments, Hoffman's benchmark is $1B annual transactional volume. For professional networks, it is the density at which your sector is well-represented. The number does not need to be precise — it needs to be directionally correct.WarningDefining critical mass as 'when the product feels good' is not a theory — it's a hope. Specify a measurable threshold.
- Build the path model from explorer behaviour to critical massMap the mechanism: how does explorer engagement drive network growth? What do explorers do that creates the signals that attract the next cohort? What is the feedback loop between early network density and new user acquisition rates? The model should be specific enough to generate testable predictions.Pro tipThe model should explain why the current majority non-use is expected. If your model only explains behaviour after critical mass, it is not a theory-of-the-game — it is a theory-of-the-endgame.
- Define leading indicators that the theory is correctWhat data patterns, if observed during the exploration phase, confirm that the theory is working? What patterns would indicate the theory is wrong? This distinguishes persistence (theory-consistent data, stay the course) from sunk-cost persistence (theory-inconsistent data, quit).Pro tipExplorer engagement rates, cohort retention among the identified explorer profile, and network density in the explorer's sector are typically more predictive leading indicators than overall MAU or revenue.
- Update the theory as you learn, but distinguish updates from abandonmentThe theory should update as you accumulate evidence — the explorer cohort might look different than expected, the critical mass threshold might be lower or higher, the feedback loop mechanism might work differently. Distinguish theory updates (revising a component based on evidence) from theory abandonment (the data pattern is inconsistent with any version of the theory).WarningTheory abandonment in response to exploration-phase majority non-use is a common mistake. The majority non-use was predicted by the theory — it is not evidence against the theory.
The majority of LinkedIn's first-wave users behaved as smart critics predicted: they did not engage meaningfully with the professional profile product. Hoffman's theory predicted this — he had modelled that 10–100 of every 1,000 early users would be explorers who saw the network's potential and that their behaviour, not the majority's, would drive the network toward critical mass.
Every prior AI wave had failed to deliver on its promises, giving smart critics legitimate grounds for skepticism. OpenAI's theory-of-the-game was specific: AI capabilities require scale compute (internet + cloud + massive data centers + massive data), and that threshold had now been crossed with the transformer architecture. The theory predicted that prior AI failures were infrastructure-limited, not capability-limited.
Hoffman developed this concept through LinkedIn's early-stage experience. LinkedIn launched in May 2003 into a market where the consensus smart-person prediction was that professionals would not build public profiles because they would seem disloyal to their employer. The early data confirmed the skeptics: the vast majority of the first wave of users behaved exactly as predicted — they did not engage meaningfully with the product.
Hoffman's theory predicted this. He had a model that said: given network-effects dynamics in professional social graphs, some fraction of early users (his estimate: 10–100 of every 1,000) would be 'explorers' — people who see the network's potential before it exists and want to be early participants. Their exploration behaviour, rather than the majority behaviour, drives the product toward the critical mass at which the value becomes self-evident. The theory allowed him to interpret the majority-non-use as expected noise rather than failure signal.