Great products copy what works, then add one new idea
Mark Pincus argues that successful products follow a pattern. They master the best existing version, improve it in ways users already want, and then add one new idea. Because most new ideas fail, the winning habits are starting small, testing fast and demanding data before belief. The same logic shapes his view of social and AI, where retention and built-in distribution matter more than hype.
Takeaways
- Great products are proven, then better, then new. Pincus's framework grew out of Zynga. Proven means mastering the best existing version of a feature. Better means polished improvements every existing user would endorse. New is the one novel idea that gets people to try the product.
- Expect most new ideas to fail. Teams should test many novel ideas and assume most will die. Zynga product managers judged a social Civilization game dead on arrival over onboarding clicks. Pincus warns founders also misuse the framework to justify wrong ideas.
- Copying is a legitimate route to winning. Pincus calls copying a moral arbitrage open to people with less ego. Freeloader built on offline browsing buried in Netscape and Internet Explorer. OMGPOP took Words With Friends' turn-based system for Draw Something. Lenny notes most products are built this way.
- The more ambitious you are, the smaller you start. Facebook began as an app for Harvard students. Pincus's tribe.net failed by doing too much without picking one use case. He then built a poker game at Zynga, which let the product find its market.
- Belief needs data, and hope is not enough. Pincus defines hope as confidence without basis and belief as resting on data. A pre-launch click test for a FarmVille expansion became $19m of early access sales. He shut down do.Earth after four years and $25m.
- Retention, not virality, built Zynga. Zynga tracked Day 365 retention and an active social network metric. Pincus says the most valuable companies have the highest Day 365 retention. Zynga was partly a victim of its viral success, since its games annoyed non-players on Facebook.
- AI is not yet a platform. Pincus says app discovery has broken down and the average app installs per user per month is zero. Distribution must be built into the product from the start. He recalls Microsoft absorbing everything around it except music and games.
Pincus's bets on what comes next
- Someone will reinvent social for the agentic AI era. Social has lost its adrenaline, and Net Promoter Scores fell sharply after people quit Facebook and Instagram.
- A free agentic travel agent is a service the AI labs might build, in his view. He says OpenAI and Claude differentiate more on coding than on consumer use.