Key Principle
Unpredictability, not the reward itself, drives craving. The nucleus accumbens activates during anticipation of reward, not upon receiving it (Knutson fMRI study). Skinner demonstrated that intermittent reinforcement dramatically increases behavior frequency compared to consistent reinforcement. A predictable reward teaches; an unpredictable one compels.
Three reward types cover the landscape of human motivation:
- Tribe -- Social validation: acceptance, attractiveness, importance. (Facebook likes, Stack Overflow upvotes.)
- Hunt -- Resource and information pursuit, rooted in persistence hunting. (Twitter feed scrolling, slot machines, Pinterest browsing.)
- Self -- Mastery and competence: intrinsic drive for completion and consistency. (Inbox zero, video game leveling, Codecademy progression.)
The strongest products layer multiple types. E-mail combines social obligation (tribe), career information (hunt), and inbox completion (self).
Variability is either finite (becomes predictable with use) or infinite (sustains itself through user-generated or multiplayer dynamics). This distinction predicts product longevity.
Why This Matters
Predictable rewards kill engagement. A plain feedback loop (action to predictable outcome) produces learning, not craving. Variability suppresses the brain's judgment centers and activates wanting centers -- dopamine surges during anticipation, not upon receipt. Without variability, the loop is a feature tutorial, not a habit engine.
Finite variability is a content treadmill. TV series, single-player games, and FarmVille clones all exhaust their patterns. Zynga went from 83.8M MAU in 2009 to an 80%+ stock decline by late 2012 as FarmVille's reward patterns became predictable. Products with finite variability must continuously produce new content or die.
Reward-trigger mismatch is the top failure mode of gamification. Variable rewards must align with the user's actual internal trigger, not the designer's assumption. Mahalo paid money for Q&A answers but failed because users' actual motivation was social validation, not income. Quora's upvoting system (social reward, zero money) proved more engaging.
Reactance destroys what rewards build. When products feel coercive, users rebel regardless of reward quality.
Good Examples
- Infinite variability (YouTube, Facebook, World of Warcraft): User-generated and multiplayer dynamics mean the system never runs out of novel combinations. WoW maintained 10M+ users eight years post-launch because other humans provide the unpredictability.
- Layered reward types (e-mail): Combines tribe (social obligation to respond), hunt (career-relevant information), and self (inbox-zero completion drive) into a single product loop.
- Autonomy-preserving compliance: The phrase "But you are free to accept or refuse" doubled compliance across a meta-analysis of 42 studies and 22,000+ participants. Variable rewards that feel like discovery rather than obligation sustain engagement.
Counterpoints
- Finite variability trap: Designing reward systems that depend on studio-produced content (new levels, new episodes) guarantees a production cost ceiling. Zynga's collapse is the canonical case.
- Reactance from coercion: Quora's auto-opt-in Views feature and forced calorie-tracking apps trigger a "hair-trigger response to threats to your autonomy." Users disengage or actively resist. Autonomy must be preserved or no reward system can compensate.
- Reward-trigger mismatch (bolted-on gamification): Adding points and badges without understanding the user's actual emotional driver produces surface engagement that decays rapidly. Mahalo vs. Quora demonstrates that matching the reward type to the real internal trigger matters more than the reward's magnitude.
Key Quotes
"What draws us to act is not the sensation we receive from the reward itself, but the need to alleviate the craving for that reward." -- Nir Eyal, Chapter 4
"Only by understanding what truly matters to users can a company correctly match the right variable reward to their intended behavior." -- Nir Eyal, Chapter 4
"Through consecutive Hook cycles, successful products reach their ultimate goal of unprompted user engagement, bringing users back repeatedly, without depending on costly advertising or aggressive messaging." -- Nir Eyal, Introduction
Rules of Thumb
- Design for anticipation, not satisfaction -- the craving state is the engine, not the payoff.
- Always ask: is this variability finite or infinite? If finite, budget for continuous content production or redesign toward user-generated dynamics.
- Layer multiple reward types (tribe + hunt + self) to create redundant engagement drivers.
- Audit for reactance: if any step feels mandatory or coercive, users will rebel regardless of reward quality.
- Match reward type to internal trigger -- run the 5 Whys before choosing tribe, hunt, or self.
- Preserve user autonomy explicitly; even small freedom cues ("you are free to...") dramatically increase compliance.
Related References
- The Action Phase and Fogg Behavior Model - What precedes reward
- The Investment Phase and Stored Value - What follows reward