Duolingo’s Growth Playbook: Ways to Build a Habit-Forming Product

Duolingo has been a success story since its humble beginnings in 2011, building a large user base of  500 million registered users. 

While that’s awesome, it isn’t surprising when you learn that its founder Luis Von Ahn was the one who founded the reCaptcha software back in 2007 and eventually sold it to Google in 2009. 

What’s common between Duolingo and reCaptcha is both were born out of using crowdsourcing to solve critical problems. 

Throughout a decade-long journey, there have been remarkable lessons learned from Duolingo HQ and its early employees. 

Some of which we’ll be discussing today.  

Duolingo’s product management lessons 

1. Go crowdsourcing mode when you’re frugal 

Your typical language experts didn’t do Duolingo’s translation. The users themselves did it. This was done to offer free language learning and motivate users to show what they have learnt. 

It grouped all the translations by students into one single algorithm that was highly accurate. The accuracy and precision was precise since 30 to 40 learners translated it. 

This brought down the labor cost. While initially, the app only offered Spanish and German, they grew and were able to raise $3.3 million in seed funding. 

Key Takeaways 

• Use crowdsourcing to bring varied opinions and balanced views
• Helps in bringing the overhead costs down as you don’t have to hire a specialized team full-time—-work can be given based on tasks 
• Helps reduce R&D costs—you can work on tasks that need more importance while delegating the rest 
• Community building and learning can go hand in hand—gives a sense of reciprocal learning 

As a product manager, here’s an example of how you can implement it. 

A project management tool could launch a feature request platform where users can submit ideas and vote on them. The most popular suggestions will be considered after evaluating them based on a mutual match between product roadmap and business outcomes. 

There are other ways to do it: 

• Adding a feature request board like this one from Ahrefs
• Public Q&A forums or Slack Discord community to solve queries 
• Bug bounty programs to identify bugs for rewards 

2. Increase engagement via gamification 

Learning has conventionally not been a fun activity, for the most part. Probably, one of the reasons why the completion rate for online courses is at a dismal 8%. 

Duolingo realizing this early on, implemented the following hacks:

• Experience points (XP): Points awarded for completing tasks; the tougher the activity, the higher the points

• Streak: The number of regular days a user spends on a task—the higher the better since it shows an increase in the DAU 

• Crowns: Crowns are rewards for completion of a level; the more the merrier since it can unlock extra features 

• Gems: Virtual currency rewarded to users based on accomplishing daily XP goals, completing quests, and maintaining streaks

• Leaderboards: A group of 30 learners in a cohort, from bronze to diamond; The top 10 learners move to the next leaderboard 

• Quests and Badges: Scoring a certain percentage of lessons completed (85% or higher in 4 lessons) means completing a quest and earning a badge; a bonus for referrals 

Badges

Key Takeaways

a. Focus on habit formation: When users stick to a consistent routine while enjoying it, their daily engagement will spike up. 

Forming a habit and sticking to it is even harder, but when you create activities that nudge users to stay consistent for 21 days(yup, it works)—users are motivated to keep their streak alive. 

Let’s take an example, as a product manager in charge of a fitness and workout app, here’s what you could do: 

• Trigger: Daily push notifications prompt users to log workouts
• Action: Users track their activities in the app
• Variable Reward: Users earn badges and personalized insights after logging
• Investment: Creating custom workout plans boosts user commitment

b. Empower users: Removing a hurdle faced by users can help you empower them. Even better, simplifying it is what you should aim for. 

Where traditional language lessons fail, Duolingo comes in and saves the day. It does it via bite-sized lessons that don’t hamper the workload of the brain. 

 In short, make the customer happy in the journey of accomplishing the goal. 

Here’s how a product manager at a fitness app would do it. 

• Customization: Users tailor workouts to their fitness goals
• Community: Groups allow users to share achievements and stay motivated

c. Leverage Instant gratification: After each lesson, giving immediate feedback on their performance can boost performance. People adhere to authority, so feedback from your instructor can help them learn better. 

When done right after the activity, users have a high recall. Similarly, ask for user feedback on your product immediately to avoid recall bias. 

Using the earlier fitness app example, here’s a product manager who can create instant gratification.

• Timely Reminders: Personalized notifications remind users of scheduled workouts (e.g., run 3x/week)
• Personalized Content: Notifications suggest classes or tips based on user preferences (e.g., yoga)

3. Do A/B Testing with a purpose 

A/B testing isn’t new but knowing the ‘why’ behind it is key before doing it. 

Here are some A/B testing lessons for product managers by Jorge Mazal, former CPO of Duolingo: 

a. Stop excessive testing

Don't over-test your channels, or you risk destroying them.

Duolingo was careful not to increase the number of push notifications sent to users without strong justification to avoid users opting out of the channel entirely.

b. Small changes equal compounding success

Duolingo found that minor improvements to the streak feature, such as adding a streak-saver notification, calendar views, animations, changes to streak freezes, and streak rewards, substantially improved retention.

c. Don’t rule out historical data

Look at historical A/B testing data to see if you have accidentally discovered successful mechanics.

When Duolingo decided to focus on improving its CURR retention metric, they looked back at old A/B test data and were surprised to find they had not run any tests that successfully moved the metric.

d. There’s nothing wrong with borrowing inspiration 

Duolingo improved its user engagement by adopting the league system from Zynga's FarmVille 2 game. 

However, they decided against blindly copying features and customized them as per Duolingo’s needs.

Users weren’t required to complete extra tasks to make it to the leaderboard. For the simple reason that it would complicate the learning experience.

e. Choose metrics that matter

Metrics tell you different stories about user segments—it's important to identify metrics that are closer to your specific goals. 

For example, Duolingo found that CURR, their key retention metric, had a much larger impact on DAU than on MAU. 

Key Takeaways

a. Identify your North Star Metric 

Duolingo used user segmentation and retention models to find CURR as their North Star Metric, driving their product decisions and DAU growth. 

You should analyze user data to find the key metric aligned with their goals (e.g., DAU or average time on the app).

b. Use data for decisions 

Duolingo modeled user behavior and retention rates to identify CURR. You can similarly analyze data to understand user behavior, identify areas for improvement, and make informed product decisions.

c. Prioritize impact 

Duolingo focused on improving CURR as it had the most significant effect on DAU. Start prioritizing initiatives that will have the biggest impact on your product goals.

d. Adapt from successful products

Duolingo adapted features from Zynga’s FarmVille 2. Don’t be afraid to draw inspiration from other products to find creative solutions for your own.

e. Optimize existing features 

Duolingo grew by optimizing push notifications and the streak mechanic. As a product person, you’d want to focus on improving current features instead of always chasing new ones.

f. Balance big wins with quick improvements 

Duolingo saw success with both major innovations (leaderboard) and small optimizations (streak-saver). You should balance breakthrough features with iterative improvements.

For example, you’re a product manager at a personal finance app for young professionals. You identify the "Number of Budget Categories Actively Tracked" (NBCAT) as the North Star Metric. 

Your data analysis reveals users tracking 5+ categories have 30% higher retention. 

As an improvement, you feel prioritizing the AI-powered category suggestion feature would boost the NBCAT, while borrowing the "Daily Finance Check-in" streak from fitness apps. 

You balance this with minor improvements like enhancing the existing categorization algorithm and adding one-tap recategorization. These data-driven decisions are expected to increase user engagement and retention.  

4. Ruthlessly focus on user retention  

Duolingo separated its users into seven buckets to better understand user engagement and identify effective strategies to increase daily active users (DAU). 

It borrowed retention models used by Zynga and MyFitnessPal—which categorized users based on their weekly return patterns. 

Duolingo created a modified in-depth model with seven buckets, each denoting a different level of user engagement.

Here's a breakdown of the buckets with four buckets that constitute DAU:

• New users: Users engaging with the app for the first time
• Current users: Users active on the given day and at least once in the preceding six days
• Reactivated users: Users returning after a 7-29 day absence
• Resurrected users: Users returning after 30 days or more of inactivity

Three buckets representing inactive users:

• At-risk WAU: Inactive on the day but active within the past six days
• At-risk MAU: Inactive for the past week but active within the previous 23 days 
• Dormant users: Inactive for 31 days or more

The model visualizes these segments as interconnected blocks, with arrows representing the transition rates between them, mirroring concepts like: 

• CURR (Current User Retention Rate) 
• NURR (New User Retention Rate) 
• RURR (Reactivated User Retention Rate)
• SURR (Resurrected User Retention Rate) 

This comprehensive system allows Duolingo to monitor daily user movement across these buckets.

By analyzing historical data and replicating the impact of incremental improvements in each retention rate, Duolingo discovered that CURR (Current User Retention Rate) had the most significant effect on DAU growth. 

This finding led to a strategic shift, prioritizing efforts to improve CURR over other metrics. Replacing it with a new user retention as a metric paid off.

The analysis also revealed that while some levers largely affected DAU, their effect on MAU was less visible. This was a call for separate strategies to drive overall user base growth.

Key Takeaways 

a. Understand user behavior affiliation

Users don’t act in silos. Segment them by engagement levels and track their movement across stages to see how actions impact retention. Duolingo shows that boosting one metric, like CURR, trickles down throughout the user lifecycle.

b. Use data to guide actions

Instead of intuition, rely on data models. Duolingo predicted how improving retention rates would affect DAU and MAU. This helped them focus on putting in efforts where required.

c. Challenge assumptions 

Duolingo shifted focus from new user retention to CURR when data showed it had a bigger impact on DAU.

d. Cross-functional collaboration 

Building user models and running simulations mandates teamwork between data scientists, engineers, and product managers.

Imagine you are a product manager for a music streaming app. You can implement Duolingo's method in the following way:

i. Segmentation: Create user segments based on listening frequency: New users, Daily listeners, Weekly listeners, Lapsing listeners (inactive for 1-2 weeks), Churned listeners (inactive for 3+ weeks) 

ii. Data analysis: Study historical data to understand the transition rates between these segments to identify key drivers of churn and reactivation

iii. Modeling: Build a model to mimic the impact of different initiatives, such as introducing a new recommendation algorithm. As an add-on, run a marketing campaign targeting lapsed users based on the North Star metric (e.g., total listening time).

With these steps, you can translate user data to make informed decisions driving sustainable growth.  

5. Identify new movable metrics 

As you move through different stages of growth, you need to identify new metrics when growth becomes stagnant. 

As Mazal puts it, “it’s vital to continuously seek new growth vectors and avoid over-reliance on any single metric.” 

Duolingo too looked past CURR with the retention team still focusing on it. But, the product and marketing teams find new metrics for acquisition and retention. 

Key Takeaways

a. Don't solely rely on optimizing one growth metric

While it's essential to focus on improving key metrics like CURR (Customer Retention Rate), it's crucial to avoid complacency. As a product manager, you should continue to explore new avenues for growth and user acquisition while anticipating stagnancy in existing strategies.

b. Diversify growth strategies 

With time, start exploring various growth vectors. This includes expanding internationally, building social features, accelerating content creation, influencer marketing, investing in paid user acquisition and leveraging viral platforms where your audience hangs out.

c. Resource allocation for growth

As companies mature, they should proportionally increase investment in growth initiatives. This includes expanding Product and Marketing teams to discover and capitalize on new opportunities for both user retention and acquisition.

d. Data-driven experimentation 

The success of a company comes from placing multiple bets on different growth strategies. This conveys the importance of a data-backed approach where various hypotheses are tested and resources are allocated based on the results. 

Wrapping Up

With Duolingo’s success, you must realize that everyone’s growth appetite is different. While you might have the aspiration, a lack of resources might not allow you to do it. 

In the end, you just need to fail first to learn fast.