Lenny's PodcastAlbert Cheng: Why amplifying users beats forcing virality
Through Chess.com losses and Grammarly sampled paid features for free users: retention is gold for subscriptions, explore-exploit picks the right mountain.
CHAPTERS
- 0:00 – 6:01
Growth as connecting users to product value (and why retention is everything)
Albert reframes growth as a user-centric discipline: connecting people to the real value of a product, not "metrics hacking." He and Lenny quickly surface a recurring theme across consumer subscriptions—retention is the fundamental lever that makes sustainable growth possible.
- •Growth = connecting users to value, not optimizing numbers in isolation
- •Retention is "gold" for consumer subscription businesses
- •Without retention, you’re forced into aggressive day-one monetization
- •Sets up the episode’s focus on mental models and experimentation
- 6:01 – 9:36
From classical piano to growth leadership: repetition, feedback loops, and creativity
Albert shares his childhood as a serious classical pianist (including perfect pitch) and how he transitioned into engineering and growth roles. He draws parallels between practicing music and running growth experiments: tight feedback loops, resilience through mistakes, and balancing structure with creativity.
- •Daily practice and perfect pitch shaped his learning style
- •Growth and music both rely on repetition and fast feedback
- •Comfort with mistakes builds experimentation resilience
- •Strong structure enables creative output in both domains
- 9:36 – 10:56
The Explore vs. Exploit framework: finding the right mountain, then climbing it
Albert introduces explore/exploit as a practical way to allocate growth effort—first discover promising opportunities, then systematically scale what works. He emphasizes the risk of being too scattershot (over-exploring) or stagnating via local optimization (over-exploiting).
- •Explore = identify high-potential growth surfaces; Exploit = scale and refine them
- •Over-exploration creates randomness; over-exploitation leads to saturation
- •Best teams oscillate between modes intentionally
- •Framework can be applied at both company and insight levels
- 10:56 – 15:19
Chess.com case study: turning loss moments into positive coaching (25% more reviews, 20% more subs)
Albert walks through an experiment that flipped a key assumption: most users review games after wins, not losses. By changing the loss experience to highlight "brilliant" moves and encouragement, Chess.com lifted game review usage, retention, and subscriptions—and then spread the insight across adjacent teams.
- •Data surprise: ~80% of reviews happened after wins
- •Loss flow redesigned to reduce negativity and increase encouragement
- •Results: +25% game reviews, +20% subscriptions, retention improved
- •Exploitation = share the insight broadly to improve other product areas
- 15:19 – 16:34
When to explore vs. exploit: detecting saturation with experiment pattern analysis
Albert explains how he looks across an experimentation portfolio to decide when a team is over-exploiting an area. A rise in non-significant results can signal diminishing returns and prompt a return to divergent brainstorming.
- •Use experiment “explorer” tooling to spot patterns across tests
- •Many non-significant results can indicate saturation
- •Leaders should proactively trigger renewed exploration
- •Scaling learning matters as much as individual wins
- 16:34 – 20:42
Using AI to accelerate experimentation: text-to-SQL, Slack bots, and faster prototyping
AI is used as leverage in the growth workflow, especially around analytics and iteration speed. Albert describes building/using Slack bots for text-to-SQL analysis and adopting prototyping tools that make ideas quickly tangible and testable across functions.
- •Text-to-SQL Slack bots answer ad hoc data questions automatically
- •AI lowers friction to asking questions, increasing data-driven behavior
- •Prototyping with tools like v0/Lovable speeds up idea-to-mock workflows
- •Challenges remain: tool interoperability and handoffs across functions
- 20:42 – 25:36
Grammarly’s biggest monetization win: sampling premium suggestions to free users
At Grammarly, the team discovered most free users didn’t reach paywalls because they accepted suggestions selectively. By interspersing a limited set of premium suggestions into the free experience, users better understood Grammarly’s full power—nearly doubling upgrade rates.
- •Fix #1: improve instrumentation around suggestions and paywall exposure
- •User behavior: free users pick-and-choose rather than accept-all
- •Strategy: show sampled premium suggestions to demonstrate value
- •Outcome: upgrade rates nearly doubled; "reverse trial" in real-time
- 25:36 – 29:26
Freemium vs. trials: designing the right subscription model for your product
Albert shares why freemium works well for mission-driven, word-of-mouth products and how to think about trials (including reverse trials). The core guidance: keep the primary value proposition free, then layer premium sampling thoughtfully based on the product’s lock-in and usage patterns.
- •Freemium supports broad mission reach and low-friction adoption
- •Free users can still drive value via word-of-mouth and team adoption
- •Keep core value free; provide structured “tastes” of premium features
- •Trials vary by context: reverse trials work best with stronger lock-in
- 29:26 – 32:26
Retention benchmarks that matter: D1 targets, habit retention, and activation-dependent products
Albert offers retention heuristics for consumer products and distinguishes new-user retention from “current user” habit retention. He notes that in some products like Grammarly, activation and installation quality can matter more than daily open behavior.
- •Heuristic: ~30–40% D1 retention is a solid baseline for many consumer apps
- •Most compounding comes from habitual retention of existing users
- •As companies mature, retention mechanics often dominate growth leverage
- •Grammarly differs: activation/aha moment is the long-term driver
- 32:26 – 34:35
Resurrected users as a growth lever: reactivation UX and social triggers
As user bases mature, a large pool of dormant users becomes a meaningful source of growth. Albert explains why reactivation flows deserve dedicated product thinking and shares Duolingo examples like social notifications and re-placement tests to reduce friction for returning learners.
- •Mature products accumulate huge dormant user populations
- •Reactivated users can rival new users in contribution to active base
- •Design a specific “resurrected user” experience (not just acquisition)
- •Examples: friend-based notifications, updated placement tests on return
- 34:35 – 41:42
How Duolingo, Grammarly, and Chess.com operate differently: culture, strategy, and core loops
Albert contrasts the operating systems of three iconic subscription businesses. Duolingo is process- and experimentation-intensive at high clock speed; Grammarly blends consumer PLG with an enterprise layer; Chess.com is deeply community- and passion-driven with heavy dogfooding.
- •Duolingo: tightly systematized experimentation and rapid iteration (Green Machine)
- •Grammarly: consumer + product-led sales; strategy shifts amid GenAI evolution
- •Chess.com: remote-first, chess-obsessed culture; constant dogfooding
- •Lesson: multiple operating models can produce exceptional outcomes
- 41:42 – 45:40
Brand + community as growth fuel: TikTok virality waves and product experimentation as the engine
Albert discusses how marketing and experimentation compound rather than compete. He describes how Duolingo’s mascot-driven brand created measurable acquisition spikes, and how chess’s broader cultural moments (pandemic, Queen’s Gambit, streaming) created wave dynamics that product teams must harness.
- •Marketing and growth experimentation are complementary, not oppositional
- •Duolingo tracked “how did you hear about us?” and saw major TikTok impact
- •Chess growth accelerated via cultural waves and creator ecosystems
- •Experimentation provides steady iteration; brand/community creates step changes
- 45:40 – 55:15
AI in chess products—and how AI reshapes the growth role and workflow
Albert explains chess engines’ long history and how Chess.com uses engines for analysis while using LLMs for approachable coaching and language. He then ties AI back to growth work: faster discovery, synthesis, ideation, and prototyping shrink cycle times, making exploration easier and cheaper.
- •Chess engines (e.g., Stockfish) far exceed top human ELO ratings
- •LLMs are used for explanation/personality; engines for deep evaluation
- •Don’t add AI for hype—apply the right tech to real user value
- •AI speeds the experiment cycle: summarization, ideation, prototyping
- 55:15 – 1:04:52
Running experiments at scale: tools, goals (1,000/year), and building an experimentation culture
Albert lays out pragmatic experimentation advice: start somewhere, instrument correctly, and focus on the system as much as any test. He shares Chess.com’s cultural shift from near-zero tests to hundreds per year and what it takes—executive support, visible wins, and enabling non-engineering experiments.
- •Many teams still don’t experiment—starting small matters
- •Tooling approaches: 3rd-party (e.g., Statsig) vs. in-house at scale
- •Chess.com trajectory: ~0 → 50 → 250 experiments/year; goal = 1,000
- •Culture change requires leadership buy-in, celebrated wins, and broad enablement
- 1:04:52 – 1:07:50
Gamification and habit-building: three pillars and onboarding beginners without discouragement
Albert shares a gamification model (core loop, metagame, profile) used at Duolingo and applies similar thinking to Chess.com’s beginner experience. He highlights how early losses and discouragement harm retention and describes experiments to guide new players more gently into learning.
- •Three gamification pillars: core loop, metagame, and profile investment
- •Duolingo’s loop: lessons → rewards/streak → notifications → repeat
- •Chess.com insight: beginners often lose early; losses reduce retention
- •Experiments: improved learn-to-play flows, hiding rating early, bots/coaches
- 1:07:50 – 1:10:38
Team-building lessons: hiring for high agency over domain experience (especially in the AI era)
Albert’s most counterintuitive team lesson is that high performers often aren’t the most experienced in a domain—they have agency, energy, and speed. He explains how to spot these traits via soft signals beyond interviews and why “beginner’s mind” matters as technology shifts quickly.
- •High agency + clock speed can outperform deep prior domain experience
- •Experience can become a crutch when the environment changes fast
- •Signals show up in preparation, questions asked, logistics, and references
- •Work trials can reveal agency better than purely conversational interviews
- 1:10:38 – 1:13:24
Choosing the right company size: finding a personal Goldilocks zone
Albert reflects on moving from big tech to tiny startups to mid-sized companies, and what each stage optimizes for. His preference is mid-sized, durable companies where he can influence strategy across the org while staying close to details and shipping weekly.
- •Big tech: scale, peers, and tooling—often slower execution
- •Tiny startups: speed and ownership—but grueling distribution/recruiting reality
- •Mid-sized (500–1,000): broad impact + ability to stay in the details
- •Look for dynamic inflection points, not stagnation
- 1:13:24 – 1:18:37
Failure Corner: Chariot Direct and lessons on validation, multi-sided UX, and premature PR
Albert shares a formative product failure from his time leading product at Chariot, where a dynamic routing concept didn’t work as hoped. The postmortem emphasizes avoiding “solutions in search of problems,” respecting multi-stakeholder complexity (drivers/ops), and delaying PR until customer value is validated.
- •Risk: building an “innovative” feature without a clear user problem
- •Marketplace reality: riders aren’t the only users—drivers and ops matter deeply
- •Premature PR creates sunk cost and pressure to persist
- •Lesson carried forward: validate via learning and iteration before hype
- 1:18:37 – 1:25:24
Lightning round: books, habits, reputation, and chess life at Chess.com
Albert closes with personal recommendations and routines—from advertising classics to a chess memoir, from espresso rituals to a reputation-focused life motto. He also shares his chess ratings, the company’s chess-friendly culture, and how to find him online and on Chess.com.
- •Book recs: Ogilvy on Advertising; Dark Squares (Danny Rensch memoir)
- •Favorite product: Breville Barista espresso machine as a daily ritual enabler
- •Life motto: protect and compound your reputation through daily choices
- •Chess.com culture includes playing at work, coaching, and constant dogfooding
