While working at Incentive Games, I designed the original product end to end — UX, UI, gameplay interaction, betting and cash-out flows, states, responsive behaviour and handoff — working closely with Light & Wonder throughout delivery. The experience was reviewed and approved by Netflix, and I then used research on V1 to challenge assumptions and improve the product before launch.
Problem
V1 worked, but player research showed it lacked entertainment value
Ownership
Designed the game end to end — UX, UI, mechanics and flows
Collaboration
Worked with Research, Product, Dev, Art, Sound, Light & Wonder and Netflix approval
Applied to
Netflix’s Squid Game: Red Light, Green Light Cashout
PROBLEM
I was responsible for designing the Red Light, Green Light experience from the ground up. The core challenge was to create a complete crash-game product that felt immediately understandable to existing players while still making the licensed Squid Game experience feel distinctive, tense and engaging.
The product needed to support the full betting journey, responsive behaviour and edge cases without allowing the interface or entertainment layer to obscure the core game mechanic.
The product problem
Create a complete, understandable betting experience that could carry the Squid Game IP while remaining clear, responsive and practical to build.
Crash games are typically built around a simple, familiar mechanic. Aviator is one of the best-known examples, using a plane and a rising multiplier to create tension with very little visual complexity.
The brief was to translate Red Light, Green Light from Netflix’s Squid Game into a crash-game format. Unlike a traditional crash game, the experience needed to feel recognisably Squid Game while still keeping the betting mechanic clear, familiar and easy to understand. The brief was intentionally open, so I had to define how the IP, gameplay and product UX should come together.
goals
Build a complete V1 that was clear enough to use and strong enough to test.
Understandable
Keep betting, cash out and game states clear even when the action becomes more dramatic.
Complete
Design the entire player journey — mechanics, flows, states, onboarding, responsive UI and handoff.
Testable
Create a real V1 that could be put in front of players and improved using evidence rather than opinion.
Testing v1
Research showed that a functional game was not enough.
After V1 was built, I worked with our full-time researcher on in-person customer research, including a visit to a William Hill shop in Newcastle. We asked players what they chose to play, why they played, what made a game feel fair and what made one experience more appealing than another.
The findings exposed a gap between the product direction and the expectations of the western audience we were targeting. Players had access to bright, animated, sound-rich experiences and described our early version as too bland by comparison. They wanted more colour, animation, sound and stronger feedback.
Research across other markets also reinforced that there was no single universal player. In African markets, familiarity, simplicity and performance on lower-powered Android devices mattered more than heavy animation or sound.
Entertainment
Western players wanted more colour, animation, sound and excitement.
Fairness
Players described live / multiplayer roulette as more trustworthy than a single-player computer simulation.
Context
African players prioritised familiar, simple experiences that performed well on lower-spec devices.
USA & UK Resarch
African Research
Evolving the direction
I used research to challenge a strong internal assumption.
There was stakeholder pushback against making the game more expressive, with a preference for keeping the experience visually restrained and avoiding sound. The research gave me evidence that this did not match what the target audience was asking for.
I used those findings to advocate for a more entertainment-led direction: stronger colour, animation, sound, anticipation and visual feedback — while keeping the betting interaction clear and familiar. The point was not to add decoration for its own sake, but to evolve V1 around observed player expectations.
This became an important product decision: we were designing for the audience using the game, not for the personal preferences of the people inside the company.
End-to-end ownership
I owned the product experience from mechanics to final UI.
This was not a redesign of somebody else’s product. I defined and designed the original game experience and then evolved it after testing. My responsibility covered the full player journey: game timing, doll-turning behaviour, normal-bet flows, Free Bet flows, cash-out flows, win and loss states, menu states, error states, How to Play, header behaviour, sound choice and responsive UI.
The most useful way to show that breadth is as one connected system of states and decisions rather than a gallery of individual screens.
Game mechanics
Normal bet
Free Bet
Cash out
Win / loss
Errors
Menu
How to Play
Header
Sound
Responsive UI
Influencing direction
Sound became a product decision, not decorative polish.
One of the clearest tensions was sound. Stakeholders initially wanted the game to remain silent, but the V1 research showed that western players expected stronger entertainment cues. I argued that we should give players control rather than remove the option based on internal preference. Netflix also required sound for the licensed experience.
Once the direction was agreed, I designed the sound-entry flow so players could choose whether to continue with sound on or off when the game loaded. I also helped identify and recommend an internal sound engineer, then worked with the wider team as audio and visual feedback were integrated into the experience.
Interaction design
Make room for the part customers came to watch.
Traditional crash layouts stacked two betting panels vertically, consuming a large part of the mobile viewport. That became a bigger problem here because the characters and game action were core to the licensed experience.
I moved the two bets side by side. This reduced the vertical footprint, kept each bet visually associated with its character and created more space for the game while retaining two simultaneous bets.
This example above shows three of our existing games side by side for comparison, although players would normally see one game at a time. The existing betting panels took up a large proportion of the available viewport even before the autoplay controls were expanded. Once autoplay was opened, the actual game area became very small. On mobile, operator compliance messaging and promotional content reduced the usable space even further.
This led me to redesign the betting panel so it used less vertical space and gave more of the screen back to the game itself.
Visual Production
Some of the hardest problems were in the characters, not the interface.
The Red Light, Green Light doll needed to feel recognisably true to the Squid Game IP while still working cleanly inside a digital product. I started with an AI-generated concept, then traced the character by hand in Illustrator so the final asset could be built as a crisp vector rather than relying on a raster image.
Netflix rejected the first version of the face, so I iterated on it repeatedly — manually adjusting individual vector points across the eyes, mouth and eyebrows until the expression was right. It took seven iterations before Netflix gave the final character the green light.
The initial stakeholder direction referenced Netflix’s Squid Game: Unleashed (shown above)as a visual benchmark. I pushed back on simply replicating an existing product and instead explored how we could create our own interpretation of Young-hee while staying recognisable to the IP.
I used AI to explore a low-poly direction quickly. After several attempts, I found a version that felt like a strong starting point for the style we wanted.
Rather than use the generated artwork directly, I rebuilt the character by hand in Illustrator as a clean vector asset. This gave me full control over the silhouette, proportions and facial details, and made the final character suitable for production.
Netflix approval became one of the more time-consuming parts of the process. Early feedback was minimal — often focused on very small details such as the eyebrows, eyes or expression — so each round required careful manual adjustments to individual vector points.
As the review process continued, the feedback became more specific. I iterated on Young-hee eleven times before the final version was approved by Netflix.
The challenge was that these changes were visually tiny but technically precise. I am not a character illustrator by trade, so manually refining the face while keeping the overall model consistent pushed me outside my usual product-design work. The final approval was a good example of persistence, attention to detail and working within strict licensed-IP constraints.
Pre-launch outcome
V1 gave us something real to test. Research gave me evidence to make it better.
I left the company before the public launch, so I do not claim post-launch performance metrics. The defensible outcome is the evolution of the product before launch: I designed the original end-to-end experience, testing exposed a mismatch between internal assumptions and player expectations, and I used that evidence to influence a more entertainment-led direction with stronger colour, animation, sound and feedback.
What this case study proves
End-to-end product ownership · UX and UI design · research-led iteration · stakeholder influence · interaction design · cross-functional delivery · responsive UX · honest handling of outcomes.
Reflection
If I were taking the project further, I would invest more deeply in Spine from the start.
One of the clearest opportunities in hindsight would have been to use Spine more broadly across the experience. It could have supported richer character motion, more expressive win states and a larger range of animation without relying on heavier frame-based approaches.
That was not on the roadmap at the time, so the team worked within the tools and constraints available. Even so, I am proud of how far the product evolved — from the complete V1 interaction and visual system through research-led changes, Netflix approval, improved sound, stronger art direction and a much more efficient character-animation approach.
What I would carry forward
Bring specialist animation capability into the project earlier, treat performance as part of the visual-design problem, and use a more scalable animation engine where the product roadmap allows it.