• Product designer
  • UX research
  • Design systems

Immersive Japanese learning inside
the anime and manga you already enjoy

A Japanese learning app where learners look up words while watching anime or reading manga, save them as scene-linked flashcards for spaced review, and choose content rated by how much of it they'll understand.

Role
Product designer, solo
Timeline
6 weeks, June–July 2025

Overview

ImmerLingo builds lessons into the shows, manga and games learners already enjoy. Watching and studying become the same thing.

Scope

  • Core loop: watch → mine → review
  • Progress tied to the media learners follow
  • Out of scope: animation and micro-interactions, music and lyrics (60% of respondents were interested).

Constraints

  • Solo and self-initiated, 6 weeks
  • Japanese only (43% of learners are learning it)
  • Dense by design for committed learners

Process

  1. 01Discovery & synthesis
  2. 02Structure & visual direction
  3. 03Design system
  4. 04Hi-fi design

Why Japanese

43% of learners surveyed are learning Japanese, the most of any foreign language, so the concept starts there.

Here's the problem it started from.

The problem

Learners quit because their apps ignore the shows that made them start.

ImageLearners describe the same problem in their own words
A Reddit User

I started learning with the hope of watching my favorite shows without subtitles, but after months of drills and memorizing random phrases, I can barely follow anything. It feels pointless when I never learn the stuff actually used in real content.

789

Full detail:5W1H breakdown, constraints and timeline→ Appendix A

Two learners showed this most clearly.

The learners

Two learners, two levels, one need: real content they can follow.

Who I designed for

Composites of 52 survey responses and 5 interviews. Split by level: 53% of learners feel unsure, like Piyush; 18% feel confident, like Orin.

Piyush Pahariya

The starter · Beginner path

  • 24 Years old
  • Pune, India
  • Part-time Barista

“I want to watch One Piece without subtitles, but I can't read hiragana yet.”

  • Absolute beginner
  • Tried apps, quit
  • Can't read kana yet
  • Planning a Japan trip

Design implication

Needs to start from zero without feeling lost.

→ Kana first, then a difficulty score that steers him to what he can follow.

Goals

  • Follow the anime he already loves, starting from zero: kana, basic words, then real episodes.
  • Pick up casual, spoken Japanese for a future trip, not textbook phrases.

Pain Points

  • Real content is too hard on day one, so he falls back to drills that feel pointless.
  • Words learned yesterday vanish, because nothing brings them back in context.
Piyush, the starter: beginner path

Orin Tyagi

The miner · Content-driven path

  • 18 Years old
  • Bangalore, India
  • Student with Tech Internship

“I want to read all the untranslated light novels and play Japanese indie games without waiting for dub patches.”

  • Knows kana
  • Reads manga & light novels
  • Plays JP indie games
  • Self-directed
  • Uses flashcards
  • Learns by reading

Design implication

Needs to grow vocabulary from what he already reads.

→ One-tap mining from video and text, with cards that keep the scene.

Goals

  • Read untranslated manga, light novels and Japanese games as they release.
  • Turn every new word he meets into one he actually remembers.

Pain Points

  • Hard to find content at his level: too easy is boring, too hard stalls him.
  • Building flashcards by hand breaks the reading flow and loses the context.
Orin, the miner: content-driven path

Where Piyush's motivation drops, and where ImmerLingo steps in

Piyush starts excited, then stalls on content he can't follow. Each dip marks where the product has to help.

  1. 01 · Excited

    Downloads anime and manga enthusiastically

    “I'm going to learn the natural way!”

  2. 02 · Delighted

    Curates perfect immersion library organized by content type

    “So much authentic content to explore!”

  3. 03 · Confused

    Opens first material, understands nothing, tries another source

    “Wait... I don't understand anything at all”

    Comprehension score before you start

  4. 04 · Exhausted

    Pauses every three seconds to look up words

    “My brain hurts from all this effort”

    Hover dictionary, no pausing

  5. 05 · Defeated

    Compares progress to others, considers quitting or switching

    “Maybe I'm just not good at languages”

    Progress tied to the series you follow

Piyush's emotional journey, and where ImmerLingo steps in

Full detail:Stage-by-stage experience flow→ Appendix D

I needed to know how common their story was.

Research

Learners want to learn through media, and the research backs them.

Four methods

A survey of 52 people, 5 interviews, a competitive audit and desk research. Each one fed the principles that follow.

Interviews

I interviewed 5 learners across Bangalore, Guwahati, Delhi, Pune and Gwalior. I asked how they practise, what frustrates them and why they quit.

Explain the numbers

Find the reasons and feelings behind the survey answers.

Find the friction

Spot the moments learners drop out of content-based learning, and why.

Test the idea

See how learners react to learning inside entertainment, and what would stop them using it.

Survey

A survey with branching logic reached 52 people across India, 38 of them active learners. It tested my assumptions and sized their needs.

68%

have no one to practise with: the #1 challenge.

71%

want tap-to-translate subtitles while watching.

73%

prefer learning words in context, from shows and books.

50%

say they forget what they learned.

1 in 2

learners feel unsure using the language they're learning.

45% lose motivation, and 29% rarely or never practise.

62%

want to learn through anime and TV shows (of all 52).

Survey of 52 people across India, 38 of them active learners

Competitors

I audited apps, browser extensions and SRS tools: what each does well, and what learners still lack.

ToolWhat it does wellWhere ImmerLingo is positioned
DuolingoStrong daily habit loop and polished onboarding for beginners.Learners in my research said drills don't transfer to real content. ImmerLingo builds practice on the shows and books they already follow.
AnkiFlexible, powerful spaced-repetition engine.Cards are built and maintained by hand. ImmerLingo creates cards at the moment of discovery, with the original sentence and scene.
Language ReactorDual subtitles and word lookup on Netflix and YouTube.Deeper review typically relies on exporting to Anki. ImmerLingo keeps lookup, mining and review in one loop.
MigakuClosest competitor: subtitle lookup, context-rich flashcards, known-word tracking and comprehension stats.Overlaps on the core loop. ImmerLingo's bet: a guided start for beginners, and progress framed around the series a learner follows.
How ImmerLingo sits against existing tools

72%

of people in India prefer watching content in other languages with subtitles.

— YouGov India (2019)

Immersion > points

Immersion-based approaches retained better long-term than point-based gamification.

— Frontiers (2024)

Time & resources

Self-directed learners struggle most with time management and finding good material.

— ResearchGate review (2024)

Three findings from published studies

Full detail:SWOT, full source quotes and published studies→ Appendix B

Insights

Motivation dies where study and entertainment split apart.

What interviews added

“I'll see a word, and then a week later, it's gone.” — Interviewee, Bangalore

Context beats drills4

  • Vocabulary sticks better in shows and books than in drills
  • Pop culture is the main reason they're learning
  • Familiar scenes make words easier to recall
  • Tools work best when they live inside the media

Motivation fades fast5

  • 45% lose motivation along the way
  • Grammar-heavy lessons feel like homework
  • Streaks build habit, not proof of progress
  • Yesterday's words are gone today
  • No one to practise with: the #1 challenge (68%)

Learning and watching are split4

  • Apps force a choice: fun or learning
  • Pausing to look up words breaks immersion
  • Saving words and reviews should happen inside the media
  • Casual and professional learners need different content
Interview findings, clustered

Context beats drills

A library of the content they already love

Learners said vocabulary sticks in shows and books, not drills. So the curriculum is anime, manga and novels, each tagged with difficulty.

Motivation fades fast

Progress that means something

Streaks kept learners consistent but didn't prove progress. Home counts words learned, heard and used, for each series they follow.

Learning and watching are split

Look up words without leaving the scene

Pausing to look things up broke immersion. A hover dictionary sits on the subtitle and resumes playback the moment you move on.

Each insight and the screen it led to

Three insights, three questions to answer.

From insight to principle

Three insights became one question and three principles.

The job, the question, the filter

Jobs to be done framed what learners want. One question framed the challenge. Three principles decided what made the cut.

The job learners hire ImmerLingo for

Learners don't hire a language app to conjugate verbs. They hire it to understand the drama, manga or game they already love.

Functional Outcome

  • Tap-to-translate and context-aware explanations for any on-screen text or dialogue.
  • Automatic vocabulary capture and review tied to specific scenes or gameplay.
  • Role-play scenarios for conversational practice in context (gaming menus, café orders, office meetings).

Emotional Outcome

  • Sustained flow state—no friction between “watch/read/play” and “learn.”
  • Confidence that progress is meaningful (not just app streaks).
  • Authentic connection to culture—feeling like a fan, traveller, or colleague rather than a student.
Jobs to be done: what learners hire ImmerLingo for

Full detail:Full friction list and complete JTBD→ Appendix C

The design challenge

How do you show detailed progress without clutter, and match every learner to content at their level?

How might we…

Each insight, reframed as a question to design against.

Learn inside the media

How might we turn watching, reading and playing into small learning moments, without breaking the flow?

Progress you can feel

How might we tie progress to episodes watched and chapters read, so milestones feel real?

Practice with others

How might we fit low-pressure speaking practice, with AI or peers, into fans' existing routines?

Keep the fandom fun

How might we keep the joy of fandom, so learning feels like being in the know, not homework?

Help only when asked

How might we show grammar tips only when learners want them, so help never feels like a lecture?

Core principles

Three principles filtered every feature.

The filter

Any feature that didn't serve at least one principle was cut.

Immersion First

Learn from real shows, manga and games, not textbook dialogues.

DrivesMedia library · Comprehension score

Active Recall

Remember words by retrieving them on a schedule, not by re-reading.

DrivesSRS deck · Scene-linked cards

Unified Experience

One system, not scattered tools. Watch, look up, mine and review in one place.

DrivesHover dictionary · One watch → mine → review loop

The evidence behind the principles

Each published finding below maps to a principle and the feature it shaped.

  1. Spacing out repeated encounters with material produces superior long-term learning; retrieval beats re-reading.

    Kang (2016) · Karpicke & Roediger (2008)

    Active Recall

    SRS deck that resurfaces each mined word just before it's forgotten, graded Again / Hard / Good / Easy.

  2. Authentic materials raised learners' observed motivation compared with artificial textbook materials.

    Peacock (1997)

    Immersion First

    A media library of anime, manga and novels as the curriculum, instead of scripted dialogues.

  3. Learners acquire language through comprehensible input pitched just beyond their current level.

    Krashen (1982), Input Hypothesis

    Immersion First

    A comprehension score on every title, benchmarked against a recommended level before the learner commits.

  4. Managing cognitive load and presenting information in small chunks improves second-language learning.

    Sweller (1988) · Mayer (2009)

    Unified Experience

    Hover dictionary inside the player: meaning, reading and one-click mining without leaving the scene.

Each finding, the principle it supports and the feature it shaped

With the principles set, I mapped the app around one loop.

Shaping the loop

One loop structures the app: watch, mine, review.

Information architecture

I organised the app around the core loop. Learning tools sit next to the content, not in separate menus.

  1. 0.0Onboarding
    1. 0.1Welcome Screen
    2. 0.2Goal Setting
    3. 0.3Skill Assessment
    4. 0.4Accessibility
    5. 0.5Schedule Setup
    6. 0.6Mascot Creator (Optional)
  2. 1.0ImmerLingo Application
    1. 7.0User Guide
    2. 8.0Community
    3. 9.0Resources
    4. 10.0Profile
  3. 1.0Language Selection
  4. 2.0Home
    1. 2.1Performance Analytics
    2. 2.3Activity
    3. 2.4Media Stats Card
    4. 2.5Voice Assist
    5. 2.6Notification
      1. 2.6.1Notification Settings
    6. 2.7Achievements
  5. 3.0Study
    1. 3.1SRS Tab
      1. 3.1.1Daily Cards
      2. 3.1.2Shared Decks
      3. 3.1.3Sync Deck
    2. 3.2Courses
      1. 3.2.1Hiragana/Katakana
      2. 3.2.2Kanji
      3. 3.2.3Grammar
    3. 3.3Converse
    4. 3.4Study Settings
  6. 4.0Clipboard
    1. 4.1Study Text
    2. 4.2OCR
  7. 5.0Word Browser
    1. 5.1Dictionary
    2. 5.2Card Creation
      1. 5.2.1Card Management
      2. 5.2.2Card Settings
  8. 6.0Settings
    1. 6.1Dictionaries
    2. 6.2Frequency List
    3. 6.3Language Setting
    4. 6.4Known Word/Sync
    5. 6.5Custom AI Prompts

Sketches

Quick iPad sketches to test layouts and interactions before committing to a structure.

ImageEarly sketches

Wireframes

Low-fidelity wireframes to check flow and hierarchy before any visual decisions.

ImageLow-fidelity wireframes

Here's that loop, through one evening with Piyush and Orin.

The experience

One evening, two learners: Piyush starting out, Orin going deeper.

Five steps, one loop

Both move through the same five steps. Piyush leans on difficulty scores and foundations; Orin on mining and review.

Step 1 of 5

Find a show at the right level

InteractiveMedia library

Pick something you can follow before you press play.

  1. Level filter

    Beginners only see what they can follow, so fewer quit early.

  2. Difficulty on every cover

    Stars and level tags show the effort before the click.

  3. Search in any script

    English, Japanese or romaji, because beginners can't type kanji yet.

“I quit last time because everything was too hard. Show me what I can follow.” — Piyush
InteractiveMedia detail

Know an episode's word load before you commit the time.

  1. Synopsis first

    Knowing the plot frees attention for the language.

  2. Difficulty panel

    Comprehension score and unknown words show the effort up front.

  3. Episode navigator

    Jump to any episode, so learning follows interest, not a syllabus.

“Tell me how many new words an episode has, then I'll decide.” — Orin
InteractiveDifficulty assessment

Decide if an episode is worth it in five seconds.

  1. Comprehension score

    Around 75% known is the sweet spot for learning from context.

  2. Known and unknown words

    95 known, 571 new: exactly what this episode will ask of you.

  3. One-new-word lines

    27% of lines have a single new word, the easiest ones to mine.

“25% means I'd be lost. I'll pick something closer to 75%.” — Piyush

Design decision

Know the difficulty before pressing play

Insight
Learners struggled to find content at their level, and 53% feel unsure in the language they're learning.
Decision
Every title shows a comprehension score and known/unknown word counts against a recommended comprehension level, before the learner commits.
Tradeoff
The score depends on known-word data. So onboarding includes a skill assessment and known-word sync from other SRS apps.

Step 2 of 5

Watch with the right subtitles

InteractivePlayer

Watch with the transcript beside you, not behind a pause button.

  1. Transcript panel

    Every line sits alongside the video, so missed lines are reread, not replayed.

  2. Bookmark a line

    Save whole sentences for grammar worth revisiting later.

  3. Score in view

    Comprehension stays visible while you watch.

“If I leave the episode to look something up, I lose the thread.” — Orin
ImageHow subtitle lookups connect to word learning
InteractiveVideo settings

Choose how much help you want, then forget the settings exist.

  1. Subtitle languages

    Japanese plus native subtitles for beginners, easy to drop later.

  2. Learning modes

    Presets for listening, reading or both, not a wall of toggles.

  3. Lookup behaviour

    Hover opens the dictionary and pauses playback. Switch it off later.

“I still need English subtitles. Let me drop them when I'm ready.” — Piyush

Step 3 of 5

Look up and mine without pausing

How mining works

Mark an unknown word while watching or reading. It becomes a flashcard with its sentence and scene, without leaving the content.

InteractiveHover dictionary

Look up a word without leaving the scene.

  1. Reading and context

    Furigana, plus where else the word shows up in what you watch.

  2. Plain meanings

    Simple definitions first, detail underneath.

  3. One-tap decision

    Known or unknown in one tap. Unknown words become cards.

“I was pausing every three seconds. I just want to hover and keep watching.” — Piyush

Design decision

Look up words without leaving the scene

Insight
“Sometimes I'll pause if I hear a word a lot and look it up, but it honestly breaks the flow.” — Interviewee, Bangalore
Decision
Hovering a subtitle word opens a popup dictionary, pauses playback and resumes on exit. One click marks the word as known or adds it to a deck.
Tradeoff
Hover is a desktop pattern. Mobile and tablet need a tap-and-hold equivalent that doesn't fight the video controls.
InteractiveMining from video

Mine a word straight from the scene.

  1. Point at the subtitle

    The word you don't know is the word you point at.

  2. Meaning in place

    The popup opens over the video. Playback waits, then resumes.

  3. Keep it

    One tap marks it known, or turns it into a card with this scene.

“Saving the word with its scene is how I'll remember it later.” — Orin
InteractiveMining from text

Manga and game text work the same way.

  1. Any text on screen

    Panels and screenshots work like subtitles, not just video.

  2. Reading above kanji

    Furigana on hover, so unfamiliar kanji don't block meaning.

  3. Same one-tap save

    Known or mined, exactly as in the player.

“Manga and game text is where I meet new words, not textbooks.” — Orin

Step 4 of 5

Review before you forget

InteractiveSRS decks

Your mined words, reviewed just before you forget them.

  1. Decks fill themselves

    Every word you mine lands in a deck, ready for review.

  2. Weekly rhythm

    Shows consistency without a streak to protect.

  3. Shared decks

    Start from community decks before you've mined your own.

“Yesterday's words vanish unless something brings them back on time.” — Piyush
ImageA word card with its original scene
ImageHow review intervals grow with each answer

Here’s a short example about how SRS works

Againyou failed → immediate repeat or short relearn.×0 or relearn step
Hardyou barely recalled → small interval increase.×1.2–1.3
Goodnormal recall → interval × Ease Factor.×EF(~2.5)
Easyeffortless recall → bigger jump in interval.×EF ×1.3 = ~3.25
  1. Sep 13 (New card)
  2. +10m
  3. +1day
  4. +3day
  5. +8day
  6. +10m
  7. +1day

*Auto-suspend cards once next review gap exceeds three months.

“I don't want to plan reviews. Just show me today's cards.” — Piyush

Step 5 of 5

See progress that means something

InteractiveHome

Home answers one question: am I improving at what I watch?

  1. Level and score

    Moves with every word you learn, hear and use, so the number means skill.

  2. Progress per series

    Each show tells you how much you understand, so you know what to watch next.

  3. A small streak

    Showing up matters, but it isn't progress, so the streak stays secondary.

“Streaks kept me opening the app. They never told me if I was improving.” — Piyush

Design decision

Progress over streaks

Insight
Streaks kept learners consistent but didn't prove progress: “Some days it feels more like checking a box than actual learning.” — Interviewee, Delhi
Decision
The dashboard leads with level progress and words learned, heard and used, tied to the series being watched or read. The streak stays as a small secondary badge.
Tradeoff
A denser screen, managed through a three-tier hierarchy: overall level, then activity, then per-series stats.
InteractiveCourses

Absolute beginners start with kana, then move to real content.

  1. Foundations first

    Hiragana, katakana and core grammar unlock everything else.

  2. Structured levels

    A clear path for beginners who aren't ready for free immersion.

  3. Community courses

    Import courses others made, so you're not stuck with defaults.

“I can't watch anything until I can read kana. Start me there.” — Piyush

Across devices

Same loop, three screens: what changes on touch.

Hover doesn't exist on a phone. These three screens show how the loop adapts from desktop to tablet to mobile.

Hover dictionary

Hover → tap-and-hold → bottom sheet

No hover on touch. A bottom sheet keeps the subtitle line in view while you read.

Desktop
Tablet
Mobile

Player

Side panel → drawer → transcript below

Screen width decides whether watching and reading share the same view.

Desktop
Tablet
Mobile

Home

Sidebar → icon rail → tab bar

Thumb reach and a single column of attention on smaller screens.

Desktop
Tablet
Mobile
How the key screens adapt across devices

Beyond the loop

Supporting tools: a word browser, text study with OCR, and settings to tailor the experience.

InteractiveWord browser

Every word you've met, one step from a flashcard.

  1. Search your words

    Find any word again, with where you first met it.

  2. Card creator

    Sentence, audio and image fields fill from the scene, or generate with AI.

  3. Create the card

    One click adds it to your deck.

“I look words up again later. Seeing where I met them saves the hunt.” — Piyush
InteractiveText study

Paste or scan any text and study it like a subtitle.

  1. Paste anything

    Lyrics, menus, game dialogue: any text becomes study material.

  2. Scan with OCR

    Signs and screenshots, where travellers meet Japanese first.

  3. Read

    Opens the text with the same hover dictionary as the player.

“Signs and menus are what I'll read in Japan. Let me scan them.” — Piyush
ImageSettings

Light mode

A light theme for daytime study.

InteractiveLight mode · Home

Light theme: the same home, readable in daylight.

  1. Same hierarchy

    Surfaces flip to light. Layout and spacing stay identical.

  2. Colour still means something

    Stat colours keep their meaning across themes.

  3. One accent

    Pink stays the accent, so actions read the same.

InteractiveLight mode · Word browser

Every component works in both themes.

  1. Tags and actions

    Tag colours and buttons hold their contrast on light surfaces.

  2. Dense forms

    Fields and labels stay readable at the same sizes.

  3. Primary action

    Create card keeps its weight in either theme.

Behind every screen is a small system that keeps it consistent.

The system underneath

A small system keeps every screen consistent.

Typography

The typeface had to render Latin and Japanese scripts with equal clarity, down to small subtitle sizes.

ImageTypography: Noto Sans and Noto Sans JP

Noto Sans

Google created Noto Sans to eliminate "tofu" boxes - those empty rectangles appearing when fonts lack specific characters, aiming to support every writing system worldwide.

Aa Bb Cc Dd Ee Ff Gg Hh Ii Jj Kk Ll Mm Nn Oo Pp Qq Rr Ss Tt Uu Vv Ww Xx Yy Zz

Noto Sans JP

あ ア, い イ, う ウ, え エ, お オ か カ, き キ, く ク, け ケ, こ コ さ サ, し シ, す ス, せ セ, そ ソ た タ, ち チ, つ ツ, て テ, と ト な ナ, に ニ, ぬ ヌ, ね ネ, の ノ は ハ, ひ ヒ, ふ フ, へ ヘ, ほ ホ ま マ, み ミ, む ム, め メ, も モ や ヤ, ゆ ユ, よ ヨ ら ラ, り リ, る ル, れ レ, ろ ロ わ ワ, を ヲ ん ン

Colour

Colours are defined in HSBA with transparency layers, so surfaces and states stay consistent in dark and light themes.

ImageColour mood

Accent

  • accent.primary
    hsba(346, 65%, 85%, 1)
    #D84B6C

Neutral

  • neutral.primary
    hsba(0, 0%, 100%, 1)
    #FFFFFF
  • neutral.secondary
    hsba(0, 0%, 5%, 1)
    #0D0D0D
  • neutral.secondary 2
    hsba(0, 0%, 5%, 0.8)
    #0D0D0D
  • neutral.grey
    hsba(0, 0%, 100%, 1)
    #FFFFFF

Tiers

  • tiers.L0
    hsba(205, 49%, 89%, 1)
    #73B4E2
  • tiers.L1
    hsba(139, 52%, 77%, 1)
    #5FC57F
  • tiers.L2
    hsba(221, 56%, 81%, 1)
    #5B80CE
  • tiers.L3
    hsba(46, 72%, 93%, 1)
    #EEC542
  • tiers.L4
    hsba(23, 70%, 97%, 1)
    #F78C49
  • tiers.L5
    hsba(346, 64%, 87%, 1)
    #DD4F71

Accent

  • accent.primary
    hsba(0, 48%, 80%, 1)
    #CD6B6B

Neutral

  • neutral.primary
    hsba(12, 19%, 11%, 1)
    #1B1716
  • Neutral/Secondary
    hsba(0, 4%, 98%, 1)
    #F9EFEF
  • Neutral/Secondary 2
    hsba(0, 0%, 100%, 0.8)
    #FFFFFF
  • Neutral/Grey
    hsba(204, 12%, 16%, 0.95)
    #232628

Tiers

  • Tiers/L0
    hsba(205, 62%, 65%, 1)
    #3F7BA6
  • Tiers/L1
    hsba(139, 82%, 53%, 1)
    #18883B
  • Tiers/L2
    hsba(221, 56%, 81%, 1)
    #153984
  • Tiers/L3
    hsba(46, 100%, 58%, 1)
    #947100
  • Tiers/L4
    hsba(23, 70%, 97%, 1)
    #CB4E00
  • Tiers/L5
    hsba(346, 68%, 81%, 1)
    #CF4263

Components

Reusable cards, buttons and inputs with defined states. Every screen is built from one set.

InteractivePopup dictionary anatomy
ImageCore components: popup dictionary and player settings

None of this has met a real learner yet. Here's how it would.

What's next

From concept to evidence.

What would decide whether this ships.

ImmerLingo hasn't been tested with learners yet. Here is the biggest risk, what I'd measure and the first test I'd run.

Biggest risk

Content access

Extracting subtitles from streaming services may conflict with platform terms and DRM. Fallbacks: YouTube captions, user-supplied subtitle files, local video and e-books.

Success metrics

What I'd track

Activation: first word mined in the first session. Retention: D7 and D30 return rate. Learning: cards reviewed per week and comprehension gained per series.

Usability test

First round

Five moderated sessions with survey respondents. Tasks: find a show at the right level, mine three words without leaving playback, and finish a review session.

Assumptions to validate

Platform accessibility

Streaming services would allow continued subtitle extraction and integration without restrictions.

Entertainment preference

Learners would choose authentic native content over traditional educational materials when given tools.

Learning commitment

Learners would put time into richer features rather than wanting a simpler tool.

Six weeks taught me a few things.

Reflection

Six weeks, one designer, and what I'd do differently.

What I'd change

I'd test earlier: the hover dictionary should have met five learners in week two. I'd design mobile first, since hover doesn't exist on a phone, and trim the dashboard for a first release.

The goal

A few months on, Piyush watches One Piece without subtitles, and Orin reads the next light novel on release day. That's the whole point.