Duolingo — improving language learning and the user’s experience

Re-designing a concept mobile app for Duolingo as a General Assembly student.

With a staggering 300 million of 1.2 billion language learners already on Duolingo, Midori Masuzawa, Eliza Edwards, Hugo Altendorf and I wanted to take some of their feedback onboard to further improve their experiences whilst learning languages and keeping it fun.

Our approach

We used the double diamond approach throughout our iterative design process.

1. Discover — Thorough research exploring the brief.
2. Define — Synthesising research into findings and the main problems.
3. Develop — Ideating towards a solution for the problems.
4. Deliver — Final design solution and future steps

The Brief

To offer quick micro-lessons and immersions into real life situations that might serve tourists, people moving abroad or family members having to integrate into a new culture quickly.

Duolingo currently teaches vocabulary and grammar linked to context, and users must reach milestones before progressing to the next stage of their learning.

The problem:

• Users are unable to learn key sentences and word for particular contexts such as ‘going to the restaurant’ or ‘inviting a girl out to dinner’.
• Users do not learn to understand cultural habits and behaviours.

Our roles

As a team we conducted competitor research, a screener survey, interviews with users of Duolingo and other language learning apps, persona creation, user journeys, users flows, information architecture, wireframes in low — high fidelity followed by prototypes. Despite the fact that we all part took in the majority of tasks, my main responsibilities were:

• Interviewing
• Persona creation
• Information architecture alteration
• Usability testing
• Iterating high fidelity prototypes

I particularly enjoyed the interview process following my recruitment of the majority of users of the Duolingo platform. I received the following feedback from an interviewee who was a UX designer and previous graduate from General Assembly:

‘Your interview style was very engaging and natural.’

Tools and software

We used Google Drive (for sharing documents), Google surveys (for the screener), Miro and Figma.

1. Discover

As a team, we all looked at various direct and indirect competitors to Duolingo. The direct competitors included Memrise, Mondly, Babbel, Rosetta Stone and Busuu.

As you can see, compared to our competitors, Duolingo sits in the less academic and less immersive corner. Their aim is to make learning enjoyable, and so we wanted to create a more immersive learning experience without losing its ethos.

By looking at certain features of the competitors shown above, it became more clear what we might like to include in our future developments. We decided to focus on the following features which would help us create a more immersive experience for our users:

• Ability to choose a scene when learning
• Pronunciation
• Cultural Highlights
• Find a native speaker

User Research

Following 24 screener survey responses, we had a good selection of users to choose from. We decided to interview 8 of the users who had used Duolingo and one or more other language learning apps to better understand the market further.

One of the main insights from the interviews was that there were two different types of users:

The casual user:

• Users would like to be able to speak and communicate whilst abroad.
• Users would like to learn to speak over any other skill.
• Users want to learn and practice in contexts.

The serious user:

• Users want to be able to chat with native speakers
• Find Duolingo too easy and would like to be challenged
• Want to learn about the culture

2. Define

Moving into the define phase, we started by creating our Personas, who emphasised the pain points experienced whilst using Duolingo currently.

This is Shelby:

We created the following How Might We (HMW) statements for Shelby and prioritised the development of the statements in bold:

HMW….

• Enable our users to practice speaking a language in different contexts
• Create a learning experience that makes our users feel motivated to continue
• Ensure our users are able to move at a pace that suits them
• Immerse our users into the country’s cultural context
• Enable users to start lessons at an appropriate level so that they do not feel like skipping

Shelby’s user journey looks like this — she starts to learn Spanish on Duolingo before going abroad, but gives up when there is a lack of contextual speaking practice to help her speak Spanish when she goes away.

This is Stella:

Stella’s user journey looked a bit like this — she starts to learn German on Duolingo to help her integrate with her parters family. She is not challenged enough in the end, and decides to choose a more challenging app.

We created the following How Might We (HMW) statements for Stella and prioritised the statements in bold as these were strongest priorities:

• Create an immersive learning experience with cultural insights and native speakers
• Keep users from choosing to go to another language learning app
• Incorporate a more academic approach to learning a new language
• Make learning and memorising words more fun

It was clear that there were different reasons that both personas might stop using the app. Shelby would lose motivation and just give up as she was not gaining enough contextual speaking practice and Stella would choose to start using an alternative app as she was not challenged enough — this needed to change!

Design Studio

From our design studio and our research, we determined that we wanted to go forward with designs for:

• A game / simulated experience to practice speaking in real situations
• Finding an effective learning partner on the app — must have similar interests

We did not have enough resources to:

• Create history of language and culture section

3. Develop

With our defined research and our design challenges in mind, we continued to produce paper sketches to begin with, tested and iterated. We then made low fidelity wireframes on Figma, which turned into mid-fidelity clickable prototypes. This was followed by testing and further iteration before moving on to high fidelity prototypes.

Paper prototypes — game to practice speaking in real situations

User testing insights:

• Too many screens
• Users got lost
• The microphone and progress buttons were not clear

Low Fidelity — Finding an Effective Learning Partner

User Testing insights:

• Confusing global navigation
• Users would message a new buddy and not call
• Prefer to play an interactive game
• Would like more information about the buddy, level, interests and age

Information Architecture

Both new paper prototype features showed that users struggled with global navigation, which confirmed insights from interviews too, so we looked at the information architecture of Duolingo and did a restructure:

Previous Information Architecture

The new information architecture included the following:

• Only 4 selective menus in the global navigation.
• ‘Connect’ contains an existing function of “Friends” and a new function of connecting language exchange partners and tutors.
• Notebook is a new function to keep track of progress, mistakes and vocabulary to remember.

Restructured Information Architecture

Mid-Fidelity Prototypes — game to practice speaking in real situations

With our new information architecture and insights from usability testing in mind, we developed mid-fidelity prototypes and tested these.

Changes the had been made in the mid-fi screens:

• Changed the context from ordering an ice cream to ordering a steak.
• Combined the menu and the practice screens for a more comprehensive experience.

Insights from Usability testing:

• Link between microphone, practice ordering and menu unclear
• Conversation dialog unclear
• Unsure when to go to next page
• ‘Finish’ button more dominant that ‘ready to practice with a buddy’
• Unsure how scoring system worked

Mid-Fidelity Prototypes — finding a study buddy

We developed mid-fidelity prototypes and tested these.

Changes the had been made in the mid-fi screens:

• Information Architecture iteration — clearer global navigation
• Added compatibility, language level, interests and age
• Remove call button

Insights from usability testing:

• Header concealed
• Users thought the names listed were suggested buddies/friends/tutors rather than current buddies/friends/tutors
• Users tried to click on the names of current buddies instead of ‘add buddy’
• Users wanted more info on tutor

‘A/B’ Testing

We conducted some testing with different users with different designs to test which ones were more practical, as screen A represented an Android home screen, and screen B represented an iPhone home screen for Duolingo. As you can see below, the streak visuals distracted the user from our new ‘learn’, ‘practice’ features, therefore we used screen B.

We presented these different versions to users, and version A kept Duolingo’s tone of voice by using the word ‘story’ and created the title ‘scene’ for the new speaking feature. This was confusing to users, therefore we developed our prototypes using version B.

4. Deliver

These are some iterations that we made from mid-high fidelity. It was now much clearer to the user what they needed to do to progress through the game to practice speaking as we made the microphone the dominant button on the practice ordering page and created onboarding with instructions.

The changes made on the study buddy iteration is displayed below. Due to confusions as to whether the list of buddies were already connected, the ‘my connections’ and ‘suggested’ were added features. Similarity was added to the profiles in order to quickly be able to scan which profile would be best suited to the user.

Hi-Fidelity prototype link:

Conclusion

To resolve the problem users were previously facing, ​our solution provided:

• A more immersive learning experience for casual learners
• Speaking practice with native speakers
• Maintaining the appealing and enjoyable ethos of the platform whilst creating better experiences for both causal and serious users

Key Learnings

• This was the first group project that we took on at General Assembly and it was a brilliant learning curve. We learnt to listen to everyones input and diplomatically make decisions based on logical insights. I thoroughly enjoyed the discussions we had to come to our final design decisions really delving deep into the problem and analysing the insights properly.
• Rearrangement of Information Architecture through site maps was another enjoyable key learning that really helped shape our final designs.
• The agile way of working amongst ourselves was so important to make sure we were on the right track and not wasting time at all times.

Future developments:

• We would like to develop the placement test that a user must partake in prior to learning a language. We believe if this placement test is accurate, the user will not want to skip to higher levels, which would omit any frustrations that users had when unable to skip levels. This would need testing.