Customer Highlight: Beating the Analytics Level at GREE

Customer Highlight: Beating the Analytics Level at GREE

Treasure Data helps a leading mobile games studio free its analysts from dependence on engineers.


GREE’s dominance in the “free-to-play” mobile gaming market has made it one of the most respected studios in the world. But analyzing user data across each of its mobile titles was anything but a game. Like most fast growing companies, data collection and analytics at GREE took a back seat to corporate strategy and product development, with titles like Knights and Dragons and War of Nations engaging 40 million active users each day.

Each title had a few loosely defined KPIs, its own custom data collection system, and a dashboard covering basic usage stats like daily user count, average play times, and so on. This was fine until executives started asking to see every game’s stats on a single dashboard. Then the data hit the fan.

A Tangled Web

Since GREE’s analytics program had grown with no oversight, none of the data collection and storage systems were compatible with each other. Each title’s raw data was locked away in its own silo, completely inaccessible to other teams. So the engineers were tasked with getting all interaction data onto a common platform. To do this, they decided to run manual scripts every 24 hours to pull data from each silo into a central Hadoop instance. This quickly became their least favorite chore, because their scripts would break every time a schema was changed. Their home grown system had turned into a game of digital Russian Roulette: open a command line, execute the script, and then find out how late you’d be working that night.

Before Treasure Data

The Case of the Missing Data

Once the datasets were centralized, an even bigger problem surfaced: none of them were complete. As an agile company, GREE’s product managers were constantly updating their games in response to user data. New kinds of events would be introduced into their analytics pipelines without any notice to engineering. The collection pipelines would silently melt down, causing permanent data loss—often at the exact time product managers were most interested in seeing the effects of their changes! Nobody was happy.

Tethered to Engineers

The final nail in the coffin was the analytics team’s dependence on engineering. Because none of GREE’s analysts new how to interact directly with Hadoop, they were always waiting on engineering to write MapReduce code to pull the datasets they needed. To make matters worse, this was hardly a one-time job. As games aged, adjustments to datasets were constantly required. For example, while mature games would have players spread evenly across hundreds of levels, brand new games saw uneven distributions of players on the lowest levels. This made standard aggregations across all titles impossible, creating even more work for the engineers, and forcing analysts to wait weeks for their data. Many of the insights that would have been useful weeks earlier went stale. GREE’s analytics program was totally on its heels.

A Surprising New Paradigm

At the height of this crisis, GREE started looking at BI tools that might give their analysts better visibility into the data without pulling it from Hadoop. One of the things they looked at was a new platform called Treasure Data. They soon realized that Treasure Data was much more than just a BI tool. It was a new paradigm for data management.

Instead of simply patching GREE’s fragmented analytics infrastructure, Treasure Data transformed it. GREE’s engineers were able to immediately replace their silos and custom ingestion scripts with SDKs that collected data seamlessly to Treasure Data’s central, managed platform. Best of all, these new pipelines were schema-flexible, allowing them to process any new format without breaking down or losing data.

Get Treasure Data blogs, news, use cases, and platform capabilities.

Thank you for subscribing to our blog!

A Complete Picture

Suddenly, GREE’s analysts had instant access to data across the entire organization. Free from having to wait weeks for engineering, they could write ad-hoc queries and aggregations by themselves on the fly. They could even send their data out to any department or analytics tool they wanted without help from engineering.

MobFox's New Analytics Architecture With Treasure DataAfter Treasure Data

With analysts in the driver’s seat, GREE’s analytics program took on a whole new life. Game-changing insights led to better user experiences and increased revenue across their entire portfolio, at a speed they’d never imagined. Treasure Data has unlocked a whole universe of new analytics possibilities at GREE, helping it entrench its dominant position and develop the most played, and playable, mobile games on the market.

Paul Lacey
Paul Lacey
Paul Lacey is equal parts entrepreneur and engineer. He manages Product Marketing at Treasure Data, after several years at a social media startup that he co-founded himself. He loves big data, ruby-on-rails, and any blogging platform other than WordPress.
Related Posts