# Educational Puzzle Game: product design case study A Number Puzzle Designed to Sharpen Quick Thinking. A number puzzle that teaches maths without ever saying so. By Krishnachandran Ramachandran (Kaezee), end-to-end product designer. Page: https://kaezee.com/work/numzy - **Product:** Numzy - **Role:** Lead UI/UX Designer - **Team:** 4 people - **Focus:** Retention Design, Multi-stakeholder Products - **Timeline:** 6+ months - **Platform:** Mobile game - **Status:** Launched - **Domain:** EdTech · a number puzzle for students and anyone who likes numbers **TL;DR.** Maths apps keep 12% of kids after a week. Puzzle games keep 60 to 70%. Numzy is a fast number puzzle: tap the numbers that fit the goal before the minute runs out. It kept 38% of players at day 7, three times the average for learning apps, and passed 1M downloads. --- ## Maths apps forgot the fun Learning apps for maths keep about 12% of kids after a week. Puzzle games like Candy Crush keep 60 to 70%. The mechanics are the same. Kids drop the app that says "learning" and cannot put down the one that says "play". **So how do you make a learning game as fun as a puzzle game?** The goal we set: 70% of players back the day after they start. **Who did what.** I was the lead designer on a team of four, and owned the design end to end, from the research to the screens. Day 7 retention maths learning app versus casual puzzle game, out of 100% Maths learning app 12% Casual puzzle game 60–70% ### Four weeks of asking I like games that make me think, not just memorize. Like Chess, but for math. **A player** Age 12 · interview - **24** Interviews, an hour each - **185** Survey answers from kids, parents and teachers - **15** Apps taken apart, learning and puzzle We took apart Khan Academy Kids, Prodigy and Monument Valley. The gap between learning apps and games was plain to see. ### Four kinds of player - **Kids, 8 to 12** Want rewards right away (figure: Kids, 8 to 12: a smiling child in a school uniform, a star badge on the chest) Bored by long explanations. - **Maths lovers** Want a real challenge (figure: Maths lovers: a smiling, freckled player in bright glasses) Want to share a clever solve. - **Teachers** Need it to match the lesson (figure: Teachers: a teacher in a cardigan holding a clipboard in one hand) Want to track progress. - **Casual gamers** Expect a polished game (figure: Casual gamers: a player in headphones holding a phone in both hands) Want short sessions. ### The four rules every feature had to pass Kids responded best to small goals they could reach straight away; big, distant goals lost them. That gave us a filter. Every feature had to pass all four rules. - Rule 1 **Fun first, maths hidden** A puzzle with no lessons and no maths words, where the maths happens on its own. - Rule 2 **Answer every action at once** Something to see, hear and feel within half a second. - Rule 3 **No tutorial** The game teaches itself. You are playing within 30 seconds of opening it. - Rule 4 **Easy to start, hard to master** Good for a 5-minute go, deep enough to keep going for months. ## Meet Numzy A fast number puzzle. Each level sets one goal, such as "multiples of 4 in ascending order". You tap the right numbers before the minute runs out. Kids build a feel for numbers without noticing they are learning. ### Try the mechanic Level **1** 01:00 Score **00** Objective Tap the **multiples of 4** in ascending order #### Decision 1: Tap to a goal, picked from five prototypes because it asks the least of you and still has depth. *The cost:* The most familiar mechanic is also the least original. Numzy had to stand out some other way. We prototyped 5 mechanics, among them tap to match, drag and drop and gesture combos. Tapping to a goal won. It asks the least of a new player and still rewards strategy. **The board never sits still.** It reshuffles after every tap, right or wrong. You cannot remember where a number was, so you read the board again each time. **Accuracy beats speed.** Every right tap is worth the same. A wrong tap costs 8 points, so only a run with no wrong taps earns three stars. *Figure: The result card: a perfect run, three wrong taps, and time's up* #### Decision 2: A square grid instead of the hex grid we loved. *The cost:* We spent the budget we had for looking different on animation polish instead. 18 students, ages 8 to 14, tested three versions. **Before, on the hex grid.** 42% finished a level. Kids were confused about how tiles moved. "It feels random," said a player, age 11. The board also grew with each difficulty, from 9 tiles to 12, 15 and 18, so the tiles shrank and the numbers got hard to read. Tiles of one colour were kept apart, and a row of one colour was hard to tap. **After, on the square grid.** 90% finished. The pattern was familiar. "Oh, I get it!" said a first-time player. Levels finished 18 students, three versions Hex grid 42% Square grid 90% *Figure: The board in four stages, drawn from the original wireframes and screens. The two tests are the playtests with 18 students.* **Bigger was not harder.** The square board grew with difficulty too, from 9 tiles to 16. Hard goals still finished fast, because the board ran out of tiles. A level had no steady length: a player could take 30 seconds on an easy level and 10 on a hard one. The numbers climbed to three digits, and sometimes four. Balancing the goal, the board and the numbers took several reworks of the engine and the difficulty curve. *Figure: The same board at three sizes. The hard board is the one from a hard level of the second test.* #### Decision 3: No tutorial, difficulty that climbs with the level, and the animation budget spent on feedback. *The cost:* A game that teaches itself has nowhere to explain the maths. Teachers lose the hook they usually get. The old tutorial took up the whole of level 1, nudging players to click through everything before they could play. For a game this simple, that was too much. Now each level opens with its goal in one sentence, and that is the teaching. Difficulty climbs level by level. We cut screen transitions and spent the time on what a tap does to a tile. *Figure: The goal card before a level starts* During play only the timer and the board carry an ink outline. Everything else steps back. *Figure: The timer, the one outlined box in the game bar, and its last ten seconds* **I took the colours through three rounds.** The first screens were all grey, and the tile colours did not stand out against it. The first test used bold, bright tiles. The second test's purple palette confused players, with white numbers on pastel tiles. In testing, patterns inside the tiles distracted players, and some found the colours too bright. What shipped is calmer: flat tiles and ink numbers. *Figure: WCAG contrast of a tile number against its tile, on a scale of 1 to 21. Tile numbers are large text, so AA needs 3:1 and AAA 4.5:1. The test colours are sampled from the test screens.* Every colour has one job. The tile colours mean nothing, so no one has to learn them. *Figure: Every colour has one job: the Numzy palette* #### Decision 4: Numzy Kids: no timer, no leaderboard, for ages 5 to 8. *The cost:* A second build to keep up, with no competition to bring players back. Numzy was built for students and anyone who likes numbers and puzzles. Numzy Kids is only for young children, ages 5 to 8. It has no timer and no leaderboard, so children build a feel for numbers without the pressure. Each level is a short set of questions, one sum at a time. *Figure: Numzy Kids: a question card with progress dots, and a control bar with no pause* ### The game, moment by moment *Figure: Right taps 12, then 24: each clears, the points go to the score, and the board reshuffles.* *Figure: A wrong tap 34 is not a multiple of 4: coral, one shake, 8 points off. Then 24, the right one.* *Figure: The last ten seconds At 00:10 the timer turns ink. At 00:00 the level ends, with no stars.* *Figure: A perfect run Four right taps, none wrong: three stars and 76 points.* #### The rest of the game: home, levels, rankings, profile, settings and Numzy Kids - (screen: Home: Play, with profile, share, rankings and settings around it) - (screen: Level select: stars earned per level, the next one open) - (screen: Rankings: your rank, then the top players) - (screen: Profile: best score, best level, badges) - (screen: Settings: music, sound and How to play) - (screen: How to play, in four steps) - (screen: Paused: the goal again, and the time it stopped at) - (screen: Numzy Kids: its own home and logo) ## After launch - **1M+** Downloads - **74%** Came back on day 1. The target was 70%. - **38%** Came back on day 7. Three times the average for learning apps. - **4.7★** Average rating, from Play Store reviews and our player communities Day 7 retention where Numzy landed, out of 100% Maths learning app 12% Numzy 38% Casual puzzle game 60–70% Did they learn? A few parents told us their child got quicker at maths. That is encouraging, but we never measured learning, so we can't claim it. My kid plays for hours without realizing he's practicing math! **A parent** Beta testing Faster than we planned Numzy launched with 100 levels. Some players finished all of them in a day, so 50 more had to follow quickly. ### What players told us, and how we prioritised Most of the raw research is lost to time. These are the playtest notes and player reviews that survived, grouped by theme. *Figure: Affinity map: playtest notes and player reviews that survived, clustered into seven themes, each with what was done about it* We sorted what we heard by how much it hurt the game and how often it came up. The praise told us what not to break. *Figure: Prioritising the feedback: how much it hurt the game against how often we heard it. Fix now: running out of levels, the steep difficulty curve, hard levels ending faster than easy ones. Fix even if quiet: shrinking hex tiles, rows of one colour. Plan for later: dark mode, a theme switcher, rewards beyond the leaderboard. Fix when cheap: bright colours, tile patterns. Protect: the simple look and the clean gameplay.* ### What players said I used to play Sudoku on my dad's old phone. This game brought those memories back. It's simple, easy to play and a good way to pass the time. **A player** Review I really love the simple UI and the clean gameplay. I love puzzle games, and this one was fun to play. **A player** Review I liked the overall feel of the game and its simple mechanics. Highly recommended if you like number puzzles. **A player** Review Numzy was retired in February 2024. ## Kids understood the familiar grid The hex grid was the idea we were proudest of, and the one kids understood least. We spent that budget on animation instead. Numzy was the company's first game, so most of this was new to all of us. - Carried forward **A leaderboard is not a reward** Players raced for the top, and all they got was their name there. The products I designed after Numzy give every player a reward with real value. - Next time **Plan the runway** 100 levels felt like plenty. Some players finished them in a day. I'd plan how new levels arrive before launch, not after. - Next time **Count up, not down** A 60-second countdown adds stress. I'd time each level instead and give stars for speed: under a set time earns three, slower earns two, and so on. It's calmer for players, and easier for the engine to scale.