Explore and Exploit for Growth

Oscillate between finding winning insights and expanding them across your product

Albert Cheng
Finding hidden growth opportunities in your product

Explore and Exploit for Growth

"When you're in exploratory mode, think of it as finding the right mountain to climb. And then when you're in exploitation mode, it's like focusing your resources on climbing that mountain effectively." - Albert Cheng

What It Is

Explore and Exploit is a framework for balancing experimentation efforts between discovery (finding new winning patterns) and optimization (expanding proven patterns across your product). Originally from reinforcement learning, Albert Cheng applies it at the insight level—not just as a macro company strategy, but as a day-to-day approach to experimentation.

The key insight: When you find something that works, don't just celebrate—systematically expand that learning across adjacent features, teams, and product areas. The winning pattern often reveals something about human psychology that applies more broadly than the specific feature tested.

How It Works

Explore Phase (Finding the Mountain)

  • Run diverse experiments across different areas
  • Test bold, divergent hypotheses
  • Look for patterns that "break through the noise"
  • Goal: Discover insights about user behavior or psychology

Exploit Phase (Climbing the Mountain)

  • Take winning insights and expand them 10X
  • Share learnings broadly across teams
  • Apply the same principle to adjacent features
  • Run variations of the winning pattern
  • Goal: Maximize the value of validated insights

Signals to Shift Phases

Time to stop exploiting (move to explore):

  • More experiments are not statistically significant
  • Diminishing returns on variations
  • Team feels "saturated" on the theme

Time to stop exploring (move to exploit):

  • A clear winning pattern emerges
  • An insight about user psychology is revealed
  • Results significantly exceed typical win rates

How to Apply It

  1. Share insights broadly - When you find a winning experiment, clearly articulate the hypothesis and what you learned. The PM who ran it doesn't need to figure out all applications—just share the insight.

  2. Encourage swarming - As a growth leader, direct multiple teams to try variations of winning patterns. "Hey, the chess review team found that positive framing after losses increased engagement 25%. How might this apply to puzzles, lessons, etc.?"

  3. Work at the insight level - Don't just track "did this experiment win?" Track "what did we learn about our users?" The insight (e.g., "users want encouragement after failure, not criticism") is more valuable than the specific implementation.

  4. Build experiment explorer tools - Invest in tooling that lets you look across experiments for patterns. AI can help surface nuggets of wisdom across large experiment volumes.

  5. Set oscillation expectations - Help teams understand they should naturally move between phases. Neither pure exploration nor pure exploitation is optimal.

Example: Chess.com Game Review

Explore phase discovery: 80% of users who review their games do so after a win, not after a loss (counterintuitive—the team expected users would want to learn from mistakes).

Insight: Users want positive reinforcement, not criticism. They engage more when the product acknowledges their good moves.

Exploit phase expansion:

  • Changed post-loss experience to highlight brilliant moves instead of blunders
  • Added encouraging coach messages ("Losing is just part of learning, keep it up")
  • Result: 25% more game reviews, 20% more subscriptions

Expanded to adjacent teams:

  • Puzzles team audited their failure states for "cold patterns"
  • Changed success ratings, copy, button colors to be more positive
  • Applied the "positivity principle" across the product

When to Use It

  • After establishing baseline experimentation (running 50+ experiments/year)
  • When you have cross-functional teams that can coordinate
  • When win rates are declining in a mature area (signal to explore)
  • When you discover a significant insight (signal to exploit)

Source

  • Guest: Albert Cheng
  • Episode: "Finding hidden growth opportunities in your product"
  • Key Discussion: (00:09:56) - Explore and exploit framework for growth teams
  • YouTube: Watch on YouTube

Related Frameworks

Back to Growth