Testing
Users feel they're being tested rather than taught, leaving lessons without feeling like they learned anything.
“Most people learn a language through actual conversations and not through short prompts or multiple choice questions.”
Student Team Project
Duolingo makes language learning approachable through short, gamified lessons. But our research found friction throughout the experience that could make it harder for users to navigate, understand, and stay engaged with their learning. We investigated where the experience broke down, validated those issues with users, and redesigned key parts of the mobile experience around what we learned.
I led the research and design process from initial heuristic review through usability testing and prototyping. I coordinated the team, analyzed research findings, and designed the improvements we ultimately tested.
PROBLEM
Duolingo’s bite-sized lessons make language learning easy to start, but our initial review revealed friction that could get in the way of learning. Users could lose their place in lessons, struggle to find key features, and encounter interruptions that pulled them away from what they were trying to accomplish.
We wanted to understand which of these issues were real user problems and which were simply artifacts of our own review. That meant putting our findings in front of people who were new to Duolingo and seeing where the experience broke down for them.

RESEARCH
We started with a heuristic review using Nielsen Norman Group’s usability heuristics as a lens. We examined the mobile experience and rated issues based on their potential impact. Rather than treating those findings as conclusions, we used them to determine what needed to be validated with real users.


The progress bar gives no indication of how many questions remain, and leaving a lesson mid-way means starting over entirely.


Language score and ELO ratings give no clear indication of progress while unlabeled icons and content leave users confused.


Duolingo is full of unskippable surveys and screens that add no personal value, slowing users down.
We recruited newly onboarded Duolingo users through a five-question screener. Nine participants then completed four moderated tasks designed around the problems we had identified during our heuristic evaluation.
Across more than 300 data points, four patterns confirmed issues from our review while another revealed a problem we had not expected.
Users feel they're being tested rather than taught, leaving lessons without feeling like they learned anything.
“Most people learn a language through actual conversations and not through short prompts or multiple choice questions.”
Users find they don't always know where they are in a lesson and when they exit their progress is lost.
“If the lessons are getting longer and I didn't know how far I was, it would bother me if it didn't save.“
Users note that it's not always easing to move through Duolingo, citing that core features can be hard to find.
“Navigating the profile is confusing. I understand what you asked me to do, I just don't know how to get there.”
Users note that frequent pop-ups break the flow of the experience and actively slow down their use of the app.
“Definitely less pop ups, because that really was annoying. I would just try to do my lesson and it was stopping me.”
The research gave us more than a list of interface problems. It helped us understand the boundaries of what this project could realistically address. The concern around practical language skills pointed toward Duolingo’s underlying learning model rather than a specific usability issue. Solving it would have required rethinking how the product teaches languages, which was beyond the scope of our redesign and the project.
DESIGN
During moderated testing, users were uncertain about where they were within a lesson. Leaving the experience could also mean losing their progress. We redesigned the lesson experience addressing both the uncertainty we observed during lessons and the frustration associated with having to start over.

Current question number is now displayed alongside progress bar.

Lesson progress now saves on exit. Users can resume from where they left off.
Another issue surfaced when users understood what Duolingo wanted them to do but couldn’t always figure out where to do it. The problem wasn’t necessarily that the feature was unclear. The path to finding it was.

Duo+ removed from navigation as it is not a core feature or user destination.

Icons now accurately reflect where they bring users without adding labels.
Our testing also showed that frequent pop-ups interrupted users while they were trying to complete lessons. Rather than responding by adding more guidance, we looked at where the interface could simply get out of the way.

Onboarding reduced to 6 core screens. Profile creation replaces survey questions.

End of lesson recap reduced down to one, skippable screen with key info.
The navigation issues were particularly relevant for newly onboarded users. We explored whether a small amount of additional guidance could help users understand the app without making onboarding more restrictive, adding an optional tutorial that introduced Duolingo’s core features and showed users where to find them.




VALIDATION
We used Maze to run an unmoderated study with eight new Duolingo users. They completed three tasks focused on areas we had redesigned. Unlike the first round of moderated testing, participants moved through these tasks without a researcher guiding them. We wanted to see how the redesigned experience performed more independently.
Of participants felt confident about their progress within lessons.
Completion rate across all tasks testing our design changes.
Less time spent on pop-ups and recaps, compared to moderated.
Faster onboarding completion compared to moderated testing.
The second round of testing gave us evidence that many participants could move through the redesigned flows successfully. It also gave us a basis for evaluating the changes we would recommend rather than relying solely on our initial design judgment.
The final recommendations focused on the improvements that addressed validated usability issues within the scope of the project. The deeper concern around Duolingo’s learning model remained outside that scope.
RECOMMENDATIONS
Our research and testing revealed opportunities to make the learning experience easier to navigate, more focused, and more flexible without changing what makes Duolingo approachable. These recommendations build on the issues we were able to validate and address within the scope of the project.
Save users’ progress within partially completed lessons and clearly show how far they are from finishing.
Use optional, contextual guidance to help users understand Duolingo’s features without forcing them through a fixed walkthrough.
Clarify where core features live and help users understand where they are within the app as they move between sections.
Remove unnecessary pop-ups, recaps, and other non-core screens that interrupt users while they are completing lessons.
Explore ways to move beyond memorization and give users more opportunities to engage with language in realistic contexts.
REFLECTION
The biggest challenge was setting our assumptions aside and letting research lead. It's easy to design for how you think users will behave, or to assume fixing a usability issue means users will respond well to the fix. What we learned is that you have to approach every test as if the user knows nothing, not as a critique but as a discipline. If I could go back, I'd build in more checkpoints to pressure test assumptions before they made it into the work.
Given more time, we'd take each insight into a focused design sprint, build out solutions, and put them in front of users again. We'd iterate on what we learned, refine the designs further, and work toward implementing changes that are as well-tested.