👋 Hey, it’s Marcus. Welcome to my newsletter segment Genesis, where I share my ideas for building upon products I love, rooted in my experience as an engineering student, AI researcher, and consumer tech product builder.
1. Inspiration
I stopped going to most of my classes this year. Even in the classes I do show up to, only about 5% of my attention is dedicated to the course content. The other 95% is checking off low-intensity to-do items, planning the rest of my day, and playing online chess.
I’ve played chess since I was young, even competing in rated tournaments during grade school. I now play to keep my mind sharp and will frequently play against my housemates during study breaks. Chess.com (if you play, add me) is one of my favorite consumer products for its low activation energy/time commitment, multitude of engagement options (bullet/blitz/rapid games, puzzles, lessons), and social features. This combination has retained me as a user for years and keeps me consistently returning to the app to play, learn, and compete.
While Duolingo has successfully gamified language learning through its lesson structure and streak system, it currently lacks a feature that is truly bite-sized, low-activation-energy, and competitive. The following case study ideates a hypothetical product to fill this gap in the Duolingo experience.
2. Opportunity
what problem are we solving?
Duolingo has established itself as a leader in language learning apps, known for its viral marketing, engaging user experience, and gamified approach to education. As the company continues to expand globally, there's a significant opportunity to enhance its offering for internationally-based English learners, particularly in China.
"Currently, China is one of the fastest-growing international markets for Duolingo. Its active users for learning English from Chinese customers have grown by over 100 percent over the past few years.”
— Rogelio Alvarez, Duolingo VP of Business & GM of Duolingo English Test
To fully capitalize on this growth, I designed a feature to introduce a new, engaging way for users to practice English - one that would complement their existing lessons and practice modules. A feature catered to learners who sought a new, skill-based challenge that would tap into their competitive spirits.
3. Users
who are we building for?
Given that English is Duolingo’s most popular language and that we are focusing on globalization, this feature should appeal to English learners in international markets.
“~50% of our users are learning English, and Spanish and French combined account for another ~25%”
— Luis von Ahn, Duolingo Co-Founder & CEO
User Persona:
“I am a motivated language learner who wants to improve my English skills for better career opportunities. I am trying to stay consistent with my lessons, but the repetitive nature of practice makes it hard to stay engaged, which makes me feel frustrated and burnt out.”
4. Product
so, what’s the solution?
Duolingo Puzzles would be a retentive freemium AI-powered puzzle feature that reinforces previously learned vocabulary through games like crosswords and word searches, increasing user engagement and retention with skill-based competition. This feature would stimulate Duolingo’s globalization and be initially intended for English learners, with potential applications for expansion into other languages.
The Practice Hub would contain a new button to access the Puzzle Hub, which would house all Puzzle-related features. Within the Puzzle Hub, users would find a suite of Puzzle features that provide extra practice and social capabilities. The free features include the Daily Crossword and a Crossword Leaderboard, much like what the NYT Games app used to offer. If the user desires more practice, they would be able to unlock unlimited crosswords, word searches, and other puzzles by subscribing to Super.
Here’s a breakdown of each of those feature groups:
Puzzle Hub
Puzzles are strategically designed as a complementary feature within the Practice Hub among the Super features, serving as another medium to reinforce skills learned in lessons. This placement acknowledges that Puzzles, while engaging and fun, are secondary to the core language learning experience.
Puzzle Generation
These Puzzles would be generated based on a user’s learned vocabulary by only including vocabulary from the user’s Section or Unit, depending on the Puzzle Type.
Puzzles would be generated internally through a sophisticated system combining:
AI-powered puzzle generation + optimized prompt engineering that ensures consistent quality and appropriate difficulty scaling.
Manual review & regular dogfooding by Duolingo team members guarantee puzzle accuracy and learning alignment.
Approved puzzles are systematically categorized into Section- or Unit-specific databases for immediate user access.
Free Features
The free tier of the Puzzle Hub is designed as a retentive mechanism to grant users a new challenge to compete every day and is defined by the Daily Crossword. By limiting free users to one puzzle per day, we create a sense of anticipation and encourage daily app visits. After playing, user completion times are recorded and added to the Daily Crossword Leaderboard to enable users to compete with their friends.
Daily Crossword
The Daily Crossword would be a single daily puzzle available to all users, generated from Section-level vocabulary (as opposed to varying at the Unit-level). This approach balances standardization with personalization by generating puzzles from users' learned vocabulary while minimizing variance and normalizing difficulty enough to maintain competitive integrity. As such, it would reinforce vocabulary specific to the user’s current learning section, enhancing knowledge retention through gamified, varied practice.
Daily Crossword Leaderboard
The Daily Crossword Leaderboard is designed with several key considerations:
Daily resets promote consistent participation and prevent long-term domination by a few high-performers.
Showing which Section's Daily Crossword a friend completed provides context for performance comparison and might motivate users to progress in their own learning to access more challenging puzzles.
Limiting the leaderboard to followed friends creates a more personal, less overwhelming competitive environment.
Users without in-app friends would be prompted to “Add Friends”, consequently promoting community building and network effects.
Premium Features
The premium tier of Puzzle Hub would amplify vocabulary retention through unlimited access to Unit-specific Puzzles, offering targeted practice tailored to each learning unit. By providing diverse puzzle types and unrestricted play, premium users can reinforce specific vocabulary sets at their own pace, complementing their core learning journey.
Game Types: Extra Practice Puzzles feature three distinct game types - Word Search, Crossword, and Connections - each generated from Unit-level vocabulary for precise learning reinforcement. This granular approach ensures users can focus on mastering specific vocabulary sets while enjoying varied puzzle mechanics. The Connections format introduces an additional cognitive layer by challenging users to identify thematic relationships between learned words, deepening their understanding of language patterns and usage.
Unlock Mechanism: As Puzzle Types get unlocked (by reaching required sections), users are presented with a step-by-step demo of how to play.
Free-To-Play: Upon clicking “Play” for Extra Practice Puzzles, Free-To-Play users would be prompted to purchase Super Duolingo.
This premium offering enables users to convert vocabulary practice from a daily event into an on-demand learning tool, allowing more motivated users to accelerate their language mastery through gamified repetition and varied challenge types.
5. KPIs
how would we define success?
To select KPIs to optimize for, we first need to understand what KPIs matter to Duolingo. Conveniently, Duolingo is structured into feature-based and metric-based teams. The former are the feature-owners that we may traditionally find in big tech companies, while the latter are solely responsible for scaling a specific growth metric that Duolingo has set its sights on.

North Star Metric: International User Engagement Rate (DAU:MAU Ratio)
I selected and prioritized engagement in particular to evaluate the retentive capacity of this product. Given the strength of Duolingo’s marketing team to launch the brand into virality, I decided to focus on cultivating strong user intensity and continuous retention to increase our DAU to MAU ratio. We also know that most of Duolingo’s revenue comes from its paid users, which account for only ~2% of its user base.
Guardrail Metric: Time Spent Learning Well (TSLW)
As we enter the solutioning phase, we want to be cautious of negatively impacting other app-wide metrics. One such metric to analyze is TSLW, which enables Duolingo to optimize for learning alongside growth.
For context, a Path lesson is any lesson node that moves you down the path, including personalized practice, Stories, and Unit Review, while Other lessons include Practice Hub exercises, level review, Side Quests, Match Madness, and Ramp Up challenges.
As Puzzles is a feature that does not reside in the Path, it is crucial to ensure that it genuinely contributes to learning outcomes rather than just entertainment. By monitoring TSLW, we can verify that time spent solving puzzles translates to meaningful vocabulary retention and doesn't detract from core learning activities.
6. Validation
how can we validate our hypothesis before investing significant design & engineering resources?
Competitors:
“Your first minimum viable product (MVP) version of an existing, widespread mechanic is not the time to get clever and innovative.”
— Sean Colombo, Duolingo VP of Engineering
The success of puzzle-based features across the digital landscape provides strong market validation for Puzzles. We can look to several established products that demonstrate the effectiveness of this model:
NYT Games' Daily Crossword - Proves the enduring appeal of daily puzzle challenges and competition within user networks
BeReal - Shows how daily rituals can become ingrained in user behavior
Chess.com - Demonstrates the power of combining free daily challenges with premium puzzle content
Our strategy deliberately mirrors these proven models with two key components:
Free Tier: A daily crossword that drives consistent engagement through competition
Premium Tier: Unlimited, AI-generated puzzles that provide flexibility and deeper practice
This approach leverages established patterns in successful puzzle-based products while tailoring them specifically for language learning. By building on proven mechanics, we can focus on optimizing the educational experience and delivering immediate value to our users.
Customers (UXR)
User Experience Research (UXR) is a critical step in product development that involves gathering feedback and insights directly from target users, allowing us to validate assumptions, understand user needs, and refine our product concept before committing significant resources to development. Here’s how I would approach UXR for Puzzles:
Create an MVP prototype with Daily Crossword and sample AI puzzles (like the User Flow above)
Design a UXR survey with Learning R&D team to assess feature interest, user sentiment, and willingness to pay for premium features
Distribute the survey to 1000-2000 Duolingo users in the target population
Implement analytics tracking in prototype to measure engagement metrics and user behavior patterns
7. Testing
Before rolling out Puzzles to all English learners, we should empirically validate our hypotheses about increased user engagement via A/B testing. By comparing different versions of the feature against a control group, we can measure its impact on the metrics defined above. This data-driven approach helps mitigate risks, reveals unexpected effects, and provides insights for optimization before a full-scale launch.
Test Design
Hypothesis:
Introducing the Puzzles feature will increase the User Engagement Rate amongst International Users by providing a new, engaging way for users to interact with the app daily.
Methodology:
Audience selection:
Target: English learners in International markets
Segment: Users who have completed at least Unit 1, Section 30 of English.
Split (Randomly assign eligible users into these two groups):
Control Group (25%): No Puzzles
Experimental Arm 1 (25%): Daily Crossword & Friends Leaderboard only
Experimental Arm 2 (25%): Premium Features only
Experimental Arm 3 (25%): Full Puzzles Suite
Rollout: Start with 4% of eligible users, increasing by 2% over 4 weeks
Tracking
Success Metrics:
During the course of the test, we would want to assess various metrics to understand the impact of this feature app-wide KPIs and engagement fluctuations with other elements of the Duolingo experience. These metrics help us quantify both the immediate impact and long-term value of the Puzzles feature, ensuring it aligns with our business objectives while enhancing the user experience.
Primary Metrics
User Engagement Rate (DAU:MAU)
Time Spent Learning Well (TSLW)
Secondary Metrics
Engagement
Daily Active Users (DAU)
Puzzle completion rate
Social interaction rate (sharing results & leaderboard checks)
Learning Effectiveness
Time spent on core learning activities
Vocabulary retention rates
Retention
Short-term (1-day), Medium-term (7-day), Long-term (30-day)
Churn rate comparison
Premium Conversion
Premium puzzle offering click-through rate
Super Duolingo conversion rate
Tracking Events
Granular event tracking allows us to understand user behavior patterns and identify potential friction points or opportunities for optimization. By collecting detailed interaction data, we can make informed decisions about feature improvements and user experience enhancements.
User Interaction Events
Puzzle starts
Puzzle completions
Time spent per puzzle
App opens
Social Events
Result shares
Leaderboard views
Friend invites
Premium Events
Premium puzzle previews
Upgrade prompts shown
Upgrade conversions
8. Conclusion
The opportunity to create engaging, retention-driving features has never been more crucial. With over 2 billion people expected to learn English by 2030, the demand for varied, bite-sized practice formats that complement traditional learning methods will only grow.
As attention spans shrink and competition for users' time intensifies, winning products are those that can transform micro-moments of idle time into meaningful opportunities to add value to users’ lives. Puzzle-based features represent more than just a new way to practice—they're a testament to how thoughtful gamification can make education more accessible, engaging, and ultimately, more human.
In an increasingly connected world where English proficiency opens doors to education and career advancement, making language learning both effective and enjoyable isn't just a product decision—it's an investment in global accessibility and opportunity.