In less than two decades, Netmarble has become one of the world’s largest mobile-gaming companies, with more than 35 titles available in 120 countries, and hit MMORPG games like Blade & Soul Revolution, Lineage 2: Revolution and most recently, BTS World. We began collaborating with Netmarble in 2017, at first to aid them in migrating to Google Cloud Platform (GCP), but more recently to help them leverage cloud tools and services to solve business challenges faced by many companies in the gaming industry.
Most recently, we’ve worked with Netmarble’s AI Center, which manages all of the company’s AI initiatives. By applying AI to their infrastructure and operations, they’ve seen a wide variety of benefits, from faster team collaboration, to more intuitive game development, to increased reach in various regions.
In this blog post, we’ll share three examples of how Netmarble and its AI Center team have worked to infuse Cloud AI into all aspects of their business, improving development, game services and operations, marketing, and player experiences overall.
ML for game services operation: churn factor analysis, churn prediction and in-game anomaly detection
For gaming companies like Netmarble with a substantial online and mobile audience, it’s not just enough to attract players, they need to retain them as well. This makes understanding why players stop playing a game–what’s known as churn–critically important.
Taking it a step further, Netmarble makes a churn prediction report, categorizing players by those who are likely to leave, who are likely to remain, and who should be managed. Based on this report, the Netmarble team can then decide each day what actions to take for each respective user group.
“The churn report has been an invaluable resource, because we hadn’t previously had access to that type of information,” said Kim. “Our next goal will be to get even more nuanced with the report, in hopes to answer tough questions like how likely are we to lose a particular player? We’ll also want to look into what preventative measures we can take to retain users who have been categorized as ‘very likely’ to abandon a game.”
One way to retain users is to continuously add new content, but this can have unintended consequences, such as increasing the number of bugs which the QA team must address. By applying machine learning for automated testing, Netmarble can quickly find any bugs–even after a high volume launch day.
As games are successfully launched, millions of users will access the game, including many fraudulent users (such as hackers and bots, for example). That’s why Netmarble uses Google Cloud AI Platform to build ML models for fraudulent user detection. Learning the growth and consumption patterns of in-game users means anomalous behavior can be quickly identified, analyzed, and aggregated into a report for further assessment.
ML in marketing: from multi-market promotions to managing ad fraud
Game marketing can be complex, with many functions to think about, such as lead generation, digital communications and game launch promotions. Additionally, Netmarble must craft strategies and battle ad fraud not in one market alone, but in both Western and Asian markets.
To address these challenges, Netmarble turned to BigQuery–a serverless, highly-scalable, and cost-effective cloud data warehouse–to build its Return On Advertising Spend (ROAS) prediction. This tool helps Netmarble predict when marketing expenses can be collected after spends in various regions. Its LTV prediction solution can evaluate the quality of users entered by cohort whom the marketer wants.
In order to cope with a variety of ad fraud challenges, Netmarble set its AD Fraud Detection system to classify heterogeneous traffic through machine learning as well as through general rule-basis as part of its detection and mediation process. This way, the company can test new media and channels preventing invalid clicks and conversions.
“In order to effectively reach the right audiences and make sure they have the best touchpoints with our games, we need to have very nuanced marketing procedures in place,” Duke Kim, SVP, Head of Netmarble’s AI Center, recently shared with us. “Google Cloud AI Platform gives us the agility and technological prowess to quickly and cost-efficiently build out internal solutions that best met these requirements.”
ML in game development: AI agents, balance checks and animation
Just as the types of games that Netmarble makes has evolved in the past 19 years, so has the way it approaches game development.
Now, the team’s next big focus is to create an AI agent that will help provide the best game experience for each player. Through this agent, which will be released soon, Netmarble intends to deliver a customized experience with specific levels, tasks or challenges tailored to a player’s skill level, that will ultimately help increase retention. This agent will offer players personalized user experiences and check how the user perceives a particular game. It will even be able to play on the user’s behalf in the event of an issue like sudden internet disconnection.
Netmarble is also looking to AI for voice and animation, which can be applied to in-game cut-scenes as well as using AI to animate the faces of its in-game non-player characters (NPCs). ML scripts prompt the NPC’s voice which then mimics mouse movements.
“I never even dreamed of some of the functionalities that AI can now bring into a game,” Kim said. “We’ve only scratched the surface of AI’s benefits for games; I’m beyond excited about what lies ahead one, five and even ten years in the future. The best part is that Google’s Cloud AI technologies have been so easy to infuse into our games, typically only taking one month. I have no doubt we’ll be able to integrate the latest AI quickly, moving forward, with Google as a collaborative partner.”
It’s been fantastic to have such a close-knit relationship with Netmarble and the entire AI Center team for the past three years. We look forward to helping them continue to reach business goals and customers in the years to come.