The main question I explored:
How can conversational AI co-create detective gameplay that enables deep thinking? I wanted to see if an AI could ask questions to enable the players to think deeply and reach conclusions logically.
I added rewarded ads on grid slots and refresh buttons, ensuring that monetization felt like an extension of the player’s decisions rather than a disruption. I analysed Bacon's Revenge item drop rate to understand how they designed their economy. I also did a weapon analysis of Doodle Magic. I took learnings from that and used it in our weapons and skill system
Here's the MDA framework that I used to fine-tune my concept
Mechanics (Designer’s Tools)
Weapon system: 4+ unique weapon archetypes (handgun, Tommy gun, shotgun, sniper, bazooka).
Progression: Mafia family ranking system where every promotion unlocks a weapon.
Clear visuals are crucial for puzzles. The player should be able to do the next action in 1-3 seconds. It was taking longer in our game. A clunky onboarding can break a game. The output should be predictable given an input. Merging game mechanics can work when they complement each other
Our mistake in the previous version was forcing a mechanic that didn’t belong. Randomly mixing genres rarely works unless the mechanics naturally reinforce the core. In this version, by focusing on the basics i.e., gun feel, enemy reactions, and simple progression, we saw a clear jump in engagement, even though retention didn’t improve much.
I think the roguelite choices were not meaningful enough. In survivor.io, there are 2 main control points of the player: Movement and Shooting. Movement is controlled by the user, and the shooting is controlled by the skill choices. It was deeply meaningful because the choices carried weight. Each choice modified the player in a significant way that impacted the power. The change was visual and striking.
To improve the powerful feeling, I added props that are destroyed when shot. I also made ammo infinite, and made it so that when shooting too much, the gun becomes hot and stuck. So users would need to stop and then shoot again when it gets close to maximum temperature. This improved the ‘power’ feeling
Ant Colony delivered on what Monument Builder lacked: lower CPI and strong early engagement. But it also exposed the limits of visual progression as a standalone driver. Without scalable monetization or deeper retention systems, the game plateaued despite one million downloads.
After analyzing hyper-casual trends, I noticed that first-person slicing games were gaining traction. From previous projects, I knew that slicing mechanics powered by third-party assets could deliver strong CPI (Cost Per Install). To add depth, I decided to integrate a bouncing disc mechanic, allowing the disc to ricochet off multiple planes and create dynamic interactions.