Duolingo - Adam Jarvis

UPLOADED

APRIL 2023

PRODUCT DESIGN

ED-TECH

A SPECULATIVE FUTURISATION OF DUOLINGO'S ED-TECH MODEL

DUOLINGO RETAINS 55% OF USERS. WHERE DID THE REST GO?

THE PROBLEM

Duolingo is one of the most prominent examples of gamified education, and since I'm passionate about learning, it's also one of my favourite products.

Duolingo's user retention rate - 45% as of 2021 - is ever improving, but in this project, I uncover and address the concerns of the remaining disengaged users.

TIMELINE

8 WEEKS · Q1 2023

PROJECT BY

ADAM JARVIS · SOLO

SET BY

SELF-INITIATED

PLATFORM

FIGMA

RESEARCHING USER NEEDS

I ASKED THEM, AND TESTED ONBOARDING.

RESEARCHING USER NEEDS

I ASKED THEM, AND TESTED ONBOARDING.

RESEARCHING USER NEEDS

I ASKED THEM, AND TESTED ONBOARDING.

RESEARCHING USER NEEDS

I ASKED THEM, AND TESTED ONBOARDING.

Upon initial review, onboarding appeared seamless and considerate of user fluency.

I talked to 41 ex-users in online language-learning communities.

I had 18 respondents to a follow-up survey, exploring the initial themes in more depth.

I remotely tested the drop-offs with 5 users - 3 dropped after 1 lesson, and 2 dropped after depleting hearts.

Upon initial review, onboarding appeared seamless and considerate of user fluency.

I talked to 41 ex-users in online language-learning communities.

I had 18 respondents to a follow-up survey, exploring the initial themes in more depth.

I remotely tested the drop-offs with 5 users - 3 dropped after 1 lesson, and 2 dropped after depleting hearts.

Upon initial review, onboarding appeared seamless and considerate of user fluency.

I talked to 41 ex-users in online language-learning communities.

I had 18 respondents to a follow-up survey, exploring the initial themes in more depth.

I remotely tested the drop-offs with 5 users - 3 dropped after 1 lesson, and 2 dropped after depleting hearts.

Upon initial review, onboarding appeared seamless and considerate of user fluency.

I talked to 41 ex-users in online language-learning communities.

I had 18 respondents to a follow-up survey, exploring the initial themes in more depth.

I remotely tested the drop-offs with 5 users - 3 dropped after 1 lesson, and 2 dropped after depleting hearts.

SOLUTION SYNTHESIS

users didn't feel A SENSE OF NATURAL LANGUAGE IMMERSION.

SOLUTION SYNTHESIS

users didn't feel A SENSE OF NATURAL LANGUAGE IMMERSION.

SOLUTION SYNTHESIS

users didn't feel A SENSE OF NATURAL LANGUAGE IMMERSION.

SOLUTION SYNTHESIS

users didn't feel A SENSE OF NATURAL LANGUAGE IMMERSION.

I found key themes among ex-users.

🤔 The more social platform alternatives are misaligned with user goals
🌱 Holistic education leads to increased user motivation
🎬 The dropped-off users often resort to movies & books to learn instead.

I followed with divergent ideation and concept validation via user journeys.

Need: be socially interactive, but not social-media.
Solve: Cross-user interaction in anonymous minigames.

Need: give proof of learning progress.
Solve: An updated fluency metric.

Need: have authentic learning materials.
Solve: TV App Media-Based Exercises.

I found key themes among ex-users.

I followed with divergent ideation and concept validation via user journeys.

Need: be socially interactive, but not social-media.
Solve: Cross-user interaction in anonymous minigames.

Need: give proof of learning progress.
Solve: An updated fluency metric.

Need: have authentic learning materials.
Solve: TV App Media-Based Exercises.

I found key themes among ex-users.

I followed with divergent ideation and concept validation via user journeys.

Need: be socially interactive, but not social-media.
Solve: Cross-user interaction in anonymous minigames.

Need: give proof of learning progress.
Solve: An updated fluency metric.

Need: have authentic learning materials.
Solve: TV App Media-Based Exercises.

I found key themes among ex-users.

I followed with divergent ideation and concept validation via user journeys.

Need: be socially interactive, but not social-media.
Solve: Cross-user interaction in anonymous minigames.

Need: give proof of learning progress.
Solve: An updated fluency metric.

Need: have authentic learning materials.
Solve: TV App Media-Based Exercises.

PROTOTYPING IT

I TESTED 'IMMERSION' BY OFFERING REAL-WORLD INTERACTIONS.

PROTOTYPING IT

I TESTED 'IMMERSION' BY OFFERING REAL-WORLD INTERACTIONS.

PROTOTYPING IT

I TESTED 'IMMERSION' BY OFFERING REAL-WORLD INTERACTIONS.

PROTOTYPING IT

I TESTED 'IMMERSION' BY OFFERING REAL-WORLD INTERACTIONS.

I drafted their implementation into Duolingo's existing architecture.

I sketched multiple wireframes, escalating the iterations that best optimised usability.

I redeveloped & expanded Duolingo's atomic design system to ensure a consistent brand experience.

Prototype feedback highlighted some issues that I was able to fix before the deadline.

Fluency points lack signposting (L)
Unclear input & visual inconsistencies (R)

I drafted their implementation into Duolingo's existing architecture.

I sketched multiple wireframes, escalating the iterations that best optimised usability.

I redeveloped & expanded Duolingo's atomic design system to ensure a consistent brand experience.

Prototype feedback highlighted some issues that I was able to fix before the deadline.

Fluency points lack signposting (L)
Unclear input & visual inconsistencies (R)

I drafted their implementation into Duolingo's existing architecture.

I sketched multiple wireframes, escalating the iterations that best optimised usability.

I redeveloped & expanded Duolingo's atomic design system to ensure a consistent brand experience.

Prototype feedback highlighted some issues that I was able to fix before the deadline.

Fluency points lack signposting (L)
Unclear input & visual inconsistencies (R)

I drafted their implementation into Duolingo's existing architecture.

I sketched multiple wireframes, escalating the iterations that best optimised usability.

I redeveloped & expanded Duolingo's atomic design system to ensure a consistent brand experience.

Prototype feedback highlighted some issues that I was able to fix before the deadline.

Fluency points lack signposting (L)
Unclear input & visual inconsistencies (R)

IMMERSING USERS IN A WORLD OF ONLINE MEDIA CONTENT

A brand-new TV App applies Duolingo's educational frameworks to media subtitles to provide realistic language exchanges.

IMMERSING USERS IN A WORLD OF ONLINE MEDIA CONTENT

A brand-new TV App applies Duolingo's educational frameworks to media subtitles to provide realistic language exchanges.

IMMERSING USERS IN A WORLD OF ONLINE MEDIA CONTENT

A brand-new TV App applies Duolingo's educational frameworks to media subtitles to provide realistic language exchanges.

DUOLINGO LEARNING + MOVIE SUBTITLES

With Duo Learning enabled, the user will be periodically assessed on their knowledge of the speech content.

I implemented a TV remote functionality into the Duolingo mobile app. This allows for a more user-friendly and accessible experience for low-vision users.

I provided users with mini-games which allow for anonymous, inter-personal communication. The games use frameworks that are adaptable to the user’s progress in Duolingo’s education structure.

Fluency points are a new user success value which can be earned through the new immersion additions. They conjoin with XP in the user profiles and leagues as ‘Points’, ensuring incentivise towards the new additions.

AFTER

USER TESTING INDICATED A 44% INCREASE IN USER RETENTION.

AFTER

USER TESTING INDICATED A 44% INCREASE IN USER RETENTION.

AFTER

USER TESTING INDICATED A 44% INCREASE IN USER RETENTION.

AFTER

USER TESTING INDICATED A 44% INCREASE IN USER RETENTION.

Feedback

I returned to the 5 test participants, and 80% said these features are likely to encourage their return to Duolingo.

Being ahead of the curve...

Duolingo MAX was unveiled soon after this project ended, using conversational AI to emulate language immersion and inter-personal exchanges, while also triggering the basis of my 2024 MSc dissertation project.

DOING THINGS DIFFERENTLY

If I was to do this project again, I would focus more on a single solution than spread across multiple. This way, the conception and testing would be more reliable and realistic.

RETURN TO PROJECTS