Project overview
Problem Research suggested that frequent practice did not always leave learners confident that they were preparing for their current goals or real-world situations.
Response I designed Personalised Practice Paths, a concept that turns an upcoming situation into a focused sequence of familiar Duolingo activities that learners can review, save and revisit.
- Role: UX/Product Designer responsible for research direction and synthesis, information architecture, interaction and UI design, prototyping, testing, iteration and final storytelling
- Scope: Two-week WBS Coding School concept project, 13–24 July 2026; not affiliated with Duolingo
- Methods: Survey, interviews, product analysis, user flows, prototyping and usability testing
- Outcome: End-to-end high-fidelity prototype and a testing-led discovery about entry-point and phrasebook findability
The project began with a brief research collaboration. I led the research direction and completed the work independently from synthesis through final delivery.
A strong habit did not always create confidence for real situations
The research included 18 survey responses across current, former and never-users, plus four interviews with current or former Duolingo users. The sample was small and directional, but it revealed an important tension. Six of seven current users practised daily or almost daily, and all seven named maintaining a streak as one reason to return.
That same system did not work identically for everyone. Streaks sustained continuity for some learners, while losing a streak or feeling pressured by it could make returning harder for others. Across current and former users, participants also described practice as repetitive, too easy, insufficiently explained or poorly matched to what they needed at that moment. Six of eight former users continued learning through other methods.
“For me, Duolingo is useful only as a day-to-day practice tool …” — Current user preparing for a B1 test with additional methods
The opportunity was therefore not to replace Duolingo’s habit loop. It was to connect that existing strength to a learner’s changing context: the trip, conversation or practical task they wanted to prepare for next.
- 18
- Survey responses
- 4
- Interviews
- 6/7
- Current users practising frequently
- 6/8
- Former users learning elsewhere
The new experience had to feel native to an established product
I translated the research into a goal-oriented path inside the existing Practice Hub. Learners could describe an upcoming situation through voice, text or a guided form. Before anything was saved, the system showed how it had interpreted the goal and which activities it proposed, allowing the learner to accept the plan or try again.
The generated path reused short, recognisable Duolingo exercise types rather than introducing a separate learning model. Learners could save it for continued practice or use it once, then return to a dedicated hub for structured activities and situation-specific Useful Phrases.
This was also a visual-systems challenge. I worked with Duolingo’s established hierarchy, shapes, colour logic, character language and concise UX writing so the new feature felt plausible within the product rather than attached to it. AI processing and voice input were simulated in Figma; the project did not test technical feasibility.
Testing uncovered a problem beyond the new feature
Three moderated mid-fidelity sessions tested path creation, plan review and access to Useful Phrases. The creation and review logic was broadly understandable in two sessions, while another participant exposed wider continuity problems between screens. The most consistent issue was the phrasebook entry point.
I had reused Duolingo’s existing placement inside a unit header. Participants overlooked it, interpreted the header as a notification or found it only through trial and error. I moved Useful Phrases into the centre of the Personalised Practice Hub’s bottom navigation, where one participant had expected a central control.
The larger high-fidelity round showed why an iteration should not be presented as a solved problem. Of ten remote participants, six generated heatmaps beyond the home-level entry point, three were explicitly recorded as completing the path and two reached the personalised phrasebook screen. Four participants never progressed beyond home-level exploration.
Moving the same icon-led pattern had changed its location without making its purpose clear. The test had surfaced a broader product question: if learners do not know that a phrasebook exists, relocation alone cannot make it discoverable.
- 10
- Remote participants
- 6
- Heatmaps beyond home
- 3
- Explicit path completions
- 2
- Phrasebook reaches
The outcome was a traceable decision chain, not a launch claim
The project resulted in an end-to-end high-fidelity prototype for a new experience that remained recognisably Duolingo. More importantly, it connected research, product framing, interaction decisions and testing without hiding where the concept still failed.
The prototype did not validate learning effectiveness, long-term retention or technical implementation. If I continued, I would:
- redesign the phrasebook entry from first principles, including its label, icon and surrounding context;
- define separate success states for entering, creating, saving and reopening a path before testing;
- validate every prototype hotspot and test with a broader range of language levels, goals and subscription tiers;
- evaluate accessibility and whether the experience improves confidence, recall or performance in a real task.
This project reflects how I prefer to work: examining the complete system, finding problems beyond the expected feature boundary and translating new behaviour into a product language people already understand.