Recent Work

I worked on Marvel Puzzle Quest after it had already built a loyal, long-running audience. The goal was to keep the game fresh for veteran players without making it harder for someone new to understand.

I led four major releases tied to Marvel theatrical moments. Each release brought together live events, new characters and mechanics, and improvements to the player experience. We compared how new, returning, and long-time players responded so we could see which changes gave people a real reason to keep playing.

I also built a P&L model for the content roadmap. It gave us a practical way to compare ideas based on development cost, expected player response, and business value. That made it easier to decide where a larger investment was justified and where a smaller update would be more responsible.

Veteran players wanted deeper strategy and worthwhile reasons to return. New players needed a clear path into a game that had accumulated years of characters, systems, and events. The same update could feel exciting to one group and confusing or too limited to the other.

We also had a full release schedule and limited development time. Every retention idea had to be specific enough for the team to understand what we would change, who it was meant to help, and how we would know whether it worked.

At first, retention was discussed as one large problem. That was not useful enough. It did not tell us which players were leaving, when they left, or what part of the experience had disappointed them.

Looking at one overall retention number also hid important differences between new and veteran players. We started breaking the audience into clearer groups and tied each release idea to a specific player problem. That gave the team a much better basis for deciding what to build.

Across these releases, Day 7 retention increased by 50% and ROAS improved by 150%. The P&L model also became a clearer way to balance creative ambition with the time and cost required to ship each update.

Marvel Puzzle Quest was named Best Puzzle Game in 2020 during this period. For me, the most important result was that the roadmap became more grounded in how different players actually experienced the game.

I worked on the global launch of AxE: Alliance vs. Empire, a mobile MMORPG that had first been designed for players in Asia. Bringing it to Western markets required more than translating the text. We had to make the game easier to learn without stripping away the depth that made it appealing.

My role covered the Western version of the product, live operations, and the go-to-market plan. We used player segments and predictive models to understand the full path from an ad or store page through onboarding, spending, and long-term play.

One of the most important decisions was when to introduce the game’s many systems. Rather than remove them, we worked to reveal them at points when players understood what they were for and were ready to make a meaningful choice.

AxE introduced progression, combat, equipment, social systems, and offers very quickly. Players familiar with the genre could move through that complexity, but many new Western players were being asked to learn too much before they understood why the game was fun.

The launch also needed the marketing and the product to tell the same story. If an ad promised one kind of experience and the first session delivered another, a strong install number would not lead to a healthy player base.

Our first pass focused too heavily on translating and explaining the existing flow. The words became clearer, but players were still meeting too many systems too early. Better instructions could not fix the order in which the game taught itself.

We changed the onboarding so complexity arrived in stages and reviewed early player behavior by audience group. We also stopped treating installs as the main measure of launch quality. Retention and long-term value became part of the same conversation from the beginning.

AxE reached 10 million downloads within six weeks, generated more than $60 million in lifetime revenue, and ranked among the top-grossing games on the Apple App Store and Google Play.

The launch also gave us a better way to judge performance across the whole player journey. We could connect the promise that brought someone into the game with the experience that kept them playing and, for some players, spending.

I supported Darkness Rises by helping the team understand where players were succeeding, struggling, or leaving. Because combat, progression, and the game economy affected one another, a problem that appeared in one place often started somewhere else.

I built internal dashboards that brought progression, completion rates, win rates, retention, and spending into one view. Instead of seeing only that players had dropped off, the team could look at what happened in the levels and decisions leading up to that point.

The dashboards became part of regular product and design reviews. That made the data useful in the room where decisions were made: we could identify a specific problem, agree on a gameplay or economy change, and then check whether it helped.

The team had overall performance numbers, but those numbers did not explain why a player stopped. A retention drop might have started with a difficulty spike, an unclear upgrade requirement, or a shortage of resources several levels earlier.

The harder part was making the analysis useful to designers. The dashboards had to follow the player’s journey and show enough detail to support a real change, not just report that a metric had moved.

At first, we relied too much on overall metrics and one-off reports. They could tell us that retention or spending had changed, but they rarely pointed to the exact moment that needed work. Broad fixes followed, and it was difficult to know which part of the change had helped.

Individual player stories were useful but could also pull attention toward unusual cases. We added repeatable views of player groups and progression steps so those stories could be checked against what the wider audience was doing.

The dashboards made it faster to move from a product question to a decision. Gameplay and economy teams had a clearer way to choose what to address first and could see whether a change improved the part of the experience it was meant to fix.

Darkness Rises went on to reach 23 million players worldwide and rank among the top-grossing RPGs on the Apple App Store and Google Play. My contribution was one part of that larger effort: helping the team make better-informed decisions as the game grew.

I was part of GREE’s data science team while War of Nations was operating at the scale of millions of players and billions of gameplay events. We had no shortage of data. The difficult part was deciding which behavior actually mattered to the health of the game.

I used regression models and grouped players by behavior to study spending, event participation, progression, and retention. The goal was to tell the difference between a temporary burst of activity and a pattern that showed players were finding lasting value in the game.

I worked with product, design, and live-operations teams to turn those findings into changes they could use. That included recommendations for event structure, progression pacing, and the way the game encouraged players to return.

A large spike during an event could mean players loved the event, that the rewards were unusually generous, or that a small group of highly active players had changed the average. Each explanation would lead to a different product decision.

Paying players were not one uniform group either. They were at different stages of the game and responded to events for different reasons. We needed to understand who was reacting before recommending what the team should change.

Overall averages made the game look healthy while hiding important differences between players. They could mask fatigue among long-time players or make a small, highly active group look representative of everyone.

One-off reports also took too long to keep up with the live event schedule. We moved toward reusable datasets and consistent player groups so the team could answer the same questions again as the game and its audience changed.

The analysis helped shape decisions about events and progression and gave the team a clearer explanation for why a result had changed. Product, design, and live operations could discuss the same player groups instead of working from different assumptions.

War of Nations became one of the studio’s highest-performing titles and ranked among the top-grossing MMO strategy games on the Apple App Store and Google Play. My role was to make player behavior easier for the team to understand and use as the game evolved.