K2 Technology
QUIZZAHH
Learning should feel like playing - not like being tested
Role
Founding Product Designer Intern
(Currently Part-time)
Team
1 founder, 2 engineers & 1 designer (me)
Duration
5 months, ongoing
Tools
Figma, Claude, Adobe Creative Cloud
Skills
UX research, Design system, Prototyping, IA, Usability testing, QA
Overview
For Vietnamese students, homework is usually a single graded shot: get it wrong, and there's no way to practice until it sticks. Teachers spend hours every day writing and marking the questions by hand. And anyone who just wants to play something smarter than a doomscroll has no good option built for Vietnam.
Quizzahh is a mobile learning quiz game that closes all three gaps at once: it generates its own questions and adapts to whatever each player gets wrong, so studying works like a game people want to return to.
As founding product designer and PM on a four-person startup, I took Quizzahh from a blank file to a pilot in three months, with two schools (~600 students) lined up once the question engine meets our quality bar.
Users
Students, Teachers, Casual players
In this project, I had one real deadline: the school year starts. Miss it and we'd miss the whole year. So the goal was never just ship an app. It was to put a study tool Vietnamese students could actually trust in their hands before the year began, and to build it lean enough that a four-person team could keep improving it once they did.
Context
Vietnam bets big on education, and its tools haven't caught up.
Education is close to a national obsession in Vietnam: families spend around a fifth of their income on it, and nearly 19 million students and teachers are in the system, almost all with a phone in hand. The appetite and the devices are there. A learning app that actually delivers is not.
0
million
students and teachers
0%
of family income spent
on education
The Market Landscape
Plenty of tools. None that teach and delight at once.
The existing options split into two camps, and each is missing half the picture.
Vietnamese Trivia Apps
Edtech Tools
Familiar and engaging
Static question banks
Ad-driven
Game over learning
—> Fun but hollow
Curriculum - aligned
Assign & grade
Unmotivating
Questions created manually
—> Serious but joyless
?
nobody here
Even the global leader leaves the gap open
Wayground reaches 90% of US schools, and still:
Gives every student the same questions: only reshuffled, no adapting to what someone actually got wrong
Isn't built for Vietnam: no localization, nothing tied to the curriculum students are tested on
Still makes teachers build the sets: a real wall for those who aren't tech-savvy
The Users
Student with one-shot practice, teacher with endless paper, and player with no good game
Katie is a strong middle-school student who learns best when she's exploring a subject herself, not when she's being drilled.
- •Actually understand a topic
- •Practice until it clicks
- •Homework is graded once: a wrong answer is a lost mark, never a lesson
- •Paper homework is static and boring
- •Loses track of what's due.
- •Mastery
- •Curiosity
- •Doing well without rote drilling
Research Synthesis
3 pains to 1 bottleneck
Making and refreshing trustworthy questions costs too much, so everyone downstream is starved of them.
One Bottleneck
Student
Practice Limited
Teacher
Homework
expensive
Quiz Player
Stale question bank
Direction framing
What If we create a core quiz flow that could serve all three?
A fast, interactive quiz flow with five question types, AI-generated and always-fresh questions, an adaptive loop that brings what users got wrong back in new forms, rankings and streaks to compete on?
Product Ideation
Making one flow for everyone
To make one core work for three different people, I gave it two doors: one for play, one for the classroom.
5 question types
all support images
AI-generated questions
always fresh
Adaptive loop
wrong answer return
Ranking & streaks
for competition
Solo or multiplayer
01
Normal Quiz Mode
Free play for anyone: game modes, rankings, streaks.
Any topic - general knowledge
Many game formats: Classic quiz, Who Wants to be a Millionaire, Road to Olympia
02
Educational Mode
Pointed at the classroom: curriculum-based questions
Homework as quizzes assigned by teacher
Self-practice by lesson
Student grading = practice process
Product Ideation
Goals that lead my design
01
Make the game teach, so a wrong answer becomes the next question instead of a lose.
02
Keep one system under everything, so new modes never fork the foundation.
03
Borrow familiar interactions, so nobody has to learn how to play.
04
Design for budget devices with less animation, because that's what most Vietnamese hold.
Success Metrics
How I know Quizzahh works
North Star for Quizzahh
Weekly concepts mastered
The number of concepts people get wrong, meet again later in a different form, and then get right, which is counted across everyone, in both quiz modes, every week. Quizzahh exists to help people actually learn, not just tap through questions. Time in app or questions answered could go up while nobody learns anything. This number only goes up when the adaptive loop does its job.
Quiz completion rate
75%
Short rounds and quick join exists to kill setup friction, so anything much lower means the flow still improving.
Question flag rate
Under 2%
If people are flagging wrong questions, nothing else matters. This is the one that kept us from rolling out to both schools on time.
Normal Quiz Mode
Quizzahh competes with other app on the phone, so retention has to be earned.
D1/D7 Retention
35%/18%
Casual games sit around 30% / 16%, and we should be at least there.
Streak length
Median streak length among returning players to tell us whether the streak mechanic is actually driving habit or is just decoration.
Educational Mode
Homework makes students show up, so we chose to watch whether they come back on their own, and whether teachers actually save time.
Teacher time to assign
Under
10 min
Teachers are busy. If this doesn’t save them real time, they’ll just go back to paper.
Unassigned
opens
A real test of whether the game part works: Do students come back when nothing is due?
Prototyping
Wireframes as the team's shared brain
With four people and a fixed deadline, the expensive mistake wasn't a bad screen. It was building the wrong thing because we each pictured it differently. So I drew the flows in lo-fi before anyone built the backend and infrastructure. Not to test on users, but to get the team and our investor looking at the same thing. Questions that are cheap to answer on a wireframe and expensive to answer in a build. Once we agreed on the skeleton, the engineers could start without waiting on me.
Prototyping
Branding and Design System
Prototyping
High fidelity for full product
Once the skeleton was agreed, I skipped straight to high fidelity. There wasn't time for anything in between. So I built the design system first: tokens, then components, then screens. Make the parts once, and assembling them gets fast.
Testing & Iterating
The main flow worked. The details didn't.
From my interview and observation, most of it worked. Navigation was easy, students found their way around within minutes, and nobody had to be taught how to play, showing the familiar patterns did their job. Still, four things got in the way, and each one changed the product:
My observation
My decision
01
Setting up a quiz took many steps for one who just wanted a quick play
Added quick join and random trivia
02
Fill-in-the-blank questions were frustrating: many different words could be right
Restricted the format to questions with exactly one answer like definitions and facts
03
Players picked an answer but ran out of time before submitting, and were marked wrong
Fixed the submission behaviour so a selected answer counts
04
New users couldn't spot where to start a quiz
Made the play button yellow, the one high-contrast target in the nav
Final Product
Normal Quiz Mode for All Players
01
Five question types, all supporting images.
Single answer
Multiple answers
True/false
Fill-in-the-blank
Word arrangement
—> Enough variety that the same knowledge can be asked five different ways, which is what keeps practice from feeling like repetition.
02
Rankings and streaks
Existing apps gave players no reason to open them twice. Now people can compete with friends and keep a streak going for every day they keep learning.
03
Daily challenge and random trivia
Testing showed the setup flow was too long for someone who just wanted to play. Two taps now gets players into a game with daily challenge and random trivia
Final Product
Educational Flow for Interactive Learning
01
Questions from the actual curriculum
The student's problem wasn't just passive practice. It was that the material they found online was scattered and sometimes incorrect. Education-mode questions are generated from the official Oxford and MOET curriculum, so what they practice matches what they are taught.
02
Homework that arrives as a quiz
Teachers spend around two hours a day writing and marking exercises, and carries the paper home. Now they review and assign from a web console, and students get it on their phone like any other game. No printing. No hand-marking!
Final Product
The Systems Underneath
Adaptive loop
The app keeps track of what each person hasn't mastered. Get something wrong, and the concept comes back later in different wording or a different question type, then again after a delay
Question engine
Every question is generated and checked by a pipeline of AI agents: one drafts from source material, one verifies, one plays the question to confirm it's answerable, one refreshes content when the underlying knowledge changes, one rewrites it to sound human
Final Thought
My Top 3 Takeaways
Three problems turned out to be one.
Once I put the interviews side by side, the same bottleneck was sitting under all three. And that's what made this something four people could actually build.
Sometimes, Lo-fi wasn't for users, it was for us.
It was the easiest way to make sure all four of us were picturing the same product.
Being the PM changed how I designed.
With four people and weeks left, I had to weigh build cost against impact and pick the five fixes worth spending the team's time on.
What I will do a bit different
Test earlier
I'd squeeze in one quick usability round between lo-fi and hi-fi, early enough to still change the structure, cheap enough to fit the deadline.
Current Status
Where is Quizzahh right now?
We aimed to launch with the school year, and we didn't make it. The engine was producing questions we weren't happy with, and putting that in front of 600 students felt worse than being late. So Quizzahh is out with a small group while we fix it, and the two schools get it once the questions hold up.
Next Step
What's Next
Formal testing
Most of my research went into understanding the problem, and the testing at the end was pretty quick and informal. Next up is a proper round: structured tasks, a real sample, things I can measure instead of just observation.
Actual numbers
Once students are using it every day, I can finally answer the questions I've been guessing at such as how many people who install it finish a quiz, how many come back tomorrow, next week, and which questions get flagged.
Shipping other quiz formats
Who Wants to Be a Millionaire and Road to Olympia are half-built. They're the formats Vietnamese players already know, which is how a new app gets noticed here.
