Playrix Boosts Game Analytics Efficiency 10X Using Amazon Redshift and AWS Fargate

As a leading casual game development company, Playrix is known for interactive mobile games. Its applications have 1.2 billion downloads and see 85 million daily active users. This creates tens of petabytes of data that the company needs to analyze to drive insights on marketing campaigns and player behavior. As player numbers continued to grow, the company’s original on-premises servers and self-hosted PostgreSQL database, along with its open-source analytics tools, struggled to scale as data quickly increased. But in the rapidly changing games industry, strong analytics are crucial to business growth and player experience. While marketing analytics are important to track the success of advertising and other campaigns to reach and acquire new players, in-game analytics are critical for tracking behavior of players, including metrics that can be used to adjust in-game mechanics. “Every successful mobile game requires not only a strong backend but also strong analytics of marketing strategies to acquire more users and in-game behaviors to optimize player retention and lifetime value,” explains Mikhail Artyugin, technical director of marketing automation at Playrix.

To first improve marketing analytics and then improve in-game analytics, Playrix—with assistance from AWS Enterprise Support—migrated to an Amazon Web Services (AWS) solution featuring Amazon Redshift, a fully managed petabyte-scale cloud data warehouse, and AWS Fargate, the engine for all extract, transform, load tasks in Playrix’s marketing analytics. Playrix uses AWS Fargate primarily for processing partner data such as ad spend, preparing data for partners, running tasks for revenue prediction, and calculating complex analytics reports that present data about the performance of marketing initiatives. Powered by the managed services of AWS, Playrix increased the speed of its analytical querying 10 times and scales to meet spikes in demand cost effectively without having an in-house database team, ultimately gaining insights that improve its business and end users’ experience.

Founded in 2004, Playrix is a mobile game developer well known for games like Township, Fishdom, Gardenscapes, Homescapes, Wildscapes, and Manor Matters, which have consistently appeared in the top 50 grossing apps in the iOS App Store and Google Play since their release. Since 2015, the company has been all in on AWS to handle the backend workloads for all its games—a transition that began in 2014, when Playrix’s first free-to-play mobile game, Township, became so popular that the company’s infrastructure couldn’t handle the load. But when the company began to struggle with performing in-game and marketing analytics, it again turned to AWS. “As our games became more popular, we received more event data,” Artyugin explains. “We need to analyze data not only for ad clicks and tens of billions of impressions per month to generate installs but also to monitor how millions of players interact with our games. We need to track feature engagement, session length per day, and more to calculate and improve player retention.”

Though Playrix had built its backend on AWS—using Amazon Elastic Compute Cloud (Amazon EC2) for compute capacity and Amazon Simple Storage Service (Amazon S3) for storage—it used an open-source SQL query engine but had problems with accuracy and the ability to modify and process current data. Playrix’s standalone PostgreSQL database was struggling to analyze more than 2 TB of event data per month. It was fragile and unscalable, and Playrix knew it was approaching the limit of its processing power. Additionally, Celery, the open-source Python task queue that Playrix used to distribute work across threads or machines, struggled to scale to handle quick increases in concurrent tasks, such as a spike from 2 to 100 tasks. But the company’s customer acquisition campaigns depended on rapid, successful data analytics. “To find new users, you have to spend money on advertising and player acquisition,” explains Artyugin. “To spend efficiently and increase success, you must analyze campaign data and create predictions to gain control over optimization.”

Playrix decided to migrate its data and analytics pipeline to Amazon Redshift partly because it’s compatible with PostgreSQL syntax and, therefore, works without overriding old SQL code. Another major factor in choosing Amazon Redshift was that, at the time, Playrix was transitioning to a new mobile attribution provider. The company wanted a service that could provide seamless trusted integrations with its partnering tools and facilitate business continuity. Playrix also decided to move away from Celery and instead, for its marketing analytics, use AWS Fargate as a compute engine, which would provide cost-effective, elastic scalability without requiring major modifications to the company’s existing code.

In 2018, Playrix began a phased approach to migrating its marketing analytics. Active migration began in the first quarter of 2020. (The migration of in-game analytics is ongoing.) Using AWS services, Playrix processes and stores up to 5 TB of data from its marketing partners. These partners push data into Amazon Kinesis Data Streams, which makes it easy to collect, process, and analyze real-time streaming data at any scale. From there, AWS Lambda filters the data and carries it into Amazon Kinesis Data Firehose, which is the easiest way to reliably load streaming data into data lakes, data stores, and analytics tools. In this case, the solution loads the data into Amazon Redshift. Playrix’s data scientists created a sophisticated machine learning model for predictions of revenue and lifetime value for users, which is deployed into AWS Fargate. “Every day we process huge amounts of data and portion the results into the dashboards and Amazon Redshift,” says Artyugin. “It’s a major part of our marketing architecture.”

Playrix achieved a highly scalable system on AWS. For example, when it needed more compute power to handle the tracking of ad clicks and the tens of billions of impressions per month that it generates, it easily upgraded from 20 Amazon Redshift DC2.xlarge nodes to three nodes of Amazon Redshift RA3.4xlarge, which enable Playrix to scale compute and storage independently for fast query performance and lower costs. Playrix uses Amazon Redshift features like Concurrency Scaling to meet dramatic spikes in workload but still keep costs low. “We set up a Concurrency Scaling feature in our cluster, and this enables us to handle spikes in the SQL queries from our internal users,” says Artyugin. During a spike in game events, Amazon Kinesis Data Streams and Amazon Kinesis Data Firehose also scale to meet the resulting spike in real-time data collection and ingest rate into Amazon Redshift.

Playrix is able to further maximize its investment by purchasing Reserved Instances for Amazon Redshift. The company also uses Savings Plans, a flexible pricing model that provides savings of up to 72 percent on its AWS compute usage, and separately, Amazon EC2 Spot Instances, which let Playrix take advantage of unused Amazon EC2 capacity in the AWS Cloud. And using AWS Fargate, Playrix is able to create a new dashboard for marketing managers, which requires quickly calculating 2–3 years of data history. “Using AWS Fargate, we can run hundreds of tasks concurrently and calculate the history in hours compared to weeks with Celery,” says Artyugin.

Since migrating to AWS, Playrix has reported a 10-times increase in the speed of its analytic querying as well as zero bottlenecks when it comes to data processing. The cost of Amazon Redshift is the same as that of its standalone PostgreSQL database running on an Amazon EC2 instance—but the ease of use, scale, and performance of Amazon Redshift is far greater. “Before it was always a hard task to stitch together infrastructure and business architecture. Now we have a very flexible architecture,” says Artyugin. “In over 3 years of using AWS, we haven’t faced a single limitation or serious barrier in implementing new features and processes.”

Because the cloud data warehouse is a managed service, Playrix doesn’t need to have a dedicated database administrator on its team and can instead put those resources toward experimenting with new features for business users. For example, Playrix can now run and support multiple versions of the same reports concurrently, add partners without worrying about the provisioning of additional resources, and quickly start loading new types of data in Amazon Redshift and analyze its value for users. Amazon Redshift also includes a feature that performs automated backups to Amazon S3 and asynchronously replicates its snapshots to Amazon S3 in another region for disaster recovery. Yet the “secret sauce” of AWS is that it enables the Playrix team to use tools that are popular with Python, database, and web-development communities: “Without vendor-specific languages or frameworks, we can apply these tools to a very flexible, very efficient, very robust infrastructure,” says Artyugin.

Playrix plans to use its AWS-powered analytics solution to stay ahead in an industry that can change in the blink of an eye. “In marketing, we’re seeing a real paradigm shift in terms of user privacy and how we should store user data, as well as the many technologies rising in mobile platforms,” says Artyugin. He specifically points to the strict privacy standards for Apple’s September 2020 iOS version: “We all need to be prepared to quickly adapt to the world where privacy rules. The companies that are not able to do so will lose a huge amount of money. So it’s very important for us to move quickly.” On AWS, Playrix has that ability to adapt quickly and introduce new capabilities. “Every feature that we want—we can build it rapidly using AWS. And we don’t need to learn something new each time. When we have a new idea or a new partner that works unusually in terms of data format, protocols, or transfers, we know that we can look in our AWS toolbox and easily build new functionality,” says Artyugin.

On AWS, Playrix built a system that scales to meet dramatic spikes in workload and process a massive amount of data to drive insights that enable Playrix to not only remain a leading competitor in the gaming industry but also improve end users’ experiences with its applications.

To learn more, visit aws.amazon.com/gametech/.