How Duolingo grew enormously from 5M to 200M users in 5 years using product-led growth.

The genius of how @duolingo grew from 5M to 200M users in 5 years (!!!) using product-led growth.

A long thread. Saddle on ๐Ÿ‘‡๐Ÿพ

1/ At volumes of millions, it’s impossible to grow sustainably without ensuring retention. Such was also the philosophy of Duolingo’s then VP Growth, @ginag.

Alongwith growing users from outside, Gina decided to double down on growing users from inside. Enter: product-led growth

2/ Duolingo’s product-led growth was based on 3 main tenets:

๐ŸŽฎ Gamification
๐Ÿ“Š Obsessive A/B testing
๐Ÿ‘ฉ๐Ÿพโ€๐ŸŽ“ Learning-driven

All 10 of their product growth-hacks fundamentally rolled back to these 3.

3/ 1โƒฃ Soft-walled sign-up
Duo had a major leak at the very onset: their sign-up page. However, it was easy to retain users once they got the Duo learning experience.

So instead, they tried this ๐Ÿ‘‡๐Ÿพ

4/

Users could begin learning instantly & faced sign-up ‘soft-walls’. These were optional sign-up screens, that could be avoided to continue learning. After ~3 softwalls, learners would lose their progress if they wouldn’t sign up.

It led to a whopping 20% increase in DAUs

5/ 2โƒฃ Lingots ๐Ÿ’Ž

Following the abc’s of any gamification handbook, Duolingo introduced their virtual currency, ‘Lingots’ onto the app (now gems). Earning them? You earned some for every lesson you completed successfully. Spending them? Lessons, perks & ‘streak freezes’.

6/ 3โƒฃ Streaks

In potentially its most powerful gamifying trick, Duo introduced ‘Streaks’. A streak is an increasing counter for the # of consecutive days a learner learns.

With every successful day, learners were rewarded to keep repeating their cognitive behavior

7/

But its real power move was in adding ‘intentioned’ goal-setting at the heart of streaks: learners set their own goals of how long they wanted to learn each day.ย 

With learners giving Duolingo a reason to call them back, Duo was now able to double down on its notifications

8/ 4โƒฃ Streak Freeze

As learners’ streaks increased, so did their willingness to retain it. To keep learners motivated, Duo launched ‘Streak freeze’.

This was a 1-day respite that users could purchase with their lingots/gems in advance.

9/ 5โƒฃ Weekend Amulets

While ‘Streak freezes’ saved you for one day, any day, weekend amulets could be purchased for that weekend specifically. Whether or not you had to use it that weekend, the amulet expired once the weekend was over.

10/

Such was the power of streaks, that people were now actively willing to spend for it, not just in virtual currency, but also in actual currency.

Streaks had dramatically increased DAUs. One of the higher streaks that’s recorded has been over 1300 days (& counting)

11/ But it wasn’t all perfect..

There was one recurring trend worrying the Duo teamโ€“the learners who lost their streaks, never got back. Gina herself quit after she lost her 80 day streak.ย 

They had to start incentivising people beyond just streaks.

12/ 6โƒฃ Badges & Leaderboard

While streaks continued engaging learners, Duolingo began visually prioritising ‘XP’ points and introduced new badges for learners.

Learners that were demotivated by streak-losses now had something else to stay engaged.

13/ 7โƒฃ Notifications

With the introduction of self-goal based streaks, Duo now had legitimate reason to support their learners in achieving their goals. With rigorous A/B tests on the copy, tone and timing of notifications, Duo optimized their notifications to the T.

14/

Notifications were optimised to be sent 23.5 hrs after the app was last used. Chances are learners would be free at the same time the next day.

Their tone was made to be friendly, more then guilt-provoking. Their effective copy tests brought a 5% increase in DAUs

15/ 8โƒฃ Notification dot on app-icon

Not just sending the notifications, but a simple hack of adding a tiny red dot to the app icon when folks had been notified to practice, showed them over 1.5% increase in DAU

16/ 9โƒฃ Content-based engagement

Having amassed an audience as large, Duo ran A/B tests on the formats of educating learnersโ€“verbs before adjectives, or vice versa?

They tested different formats to identify which engaged the learners towards the most efficient learning

17/ ๐Ÿ”Ÿ Chatbot

One of the biggest deterrents in practicing languages is the fear of embarrassing ourselves. Duo completely skipped past that when they introduced their virtual chatbot.

Learners were now able to have even amateur conversation without the inertia of embarrassment

18/ Summation

In all that Duo & @ginag had made possible, there were 2 biggest learnings they uncovered:
๐Ÿ’ก Users don’t know what they want. Feed everything to an A/B test
๐Ÿ’ก Testing is a game of instinct. In infinite option-sets, you can still realistically only test the top 3

19/ Concluding remarks

What Duolingo & @Ginag were able to achieve in 5 yrs is a strong case for the power of product-led growth.

What an absolute delight it’s been to study this genius case by an org like Doulingo, built on a clear force of purpose: accessible education


» NextBigWhat’s #Threadmill brings you curated wisdom from Twitter threads on product, life and growth.

Leave a Reply