Go3 drives personalized customer experiences with AWS data platform

With a compound annual growth rate (CAGR) nearing 30%, the over-the-top (OTT) market is bound to reach over USD 2 trillion by 2030. As more and more consumers turn to the flexibility of online streaming services, competition is fierce — especially across video streaming providers. In a market dominated by heavyweights and cloud pioneers, innovation is key as enterprises turn to providing their own OTT services.

TV3 Group — the leading Media Group in the Baltics — is among those at the forefront of this wave. The 1993-founded company delivers a wide range of media services including an array of both free and subscription-based TV channels and radio stations, also offered in AVOD/SVOD models. With over 600k subscribers, representing 20% of household penetration in the region, more than 600 live channels, and more than 20 thousand hours of content, TV3 Group's OTT service Go3, provides the most extensive internet TV service across Estonia, Latvia, and Lithuania.

Opportunity Achieving a competitive edge

OTT platforms can decode user behavior through insights derived from millions of daily interactions, and this was exactly what Go3 needed — a solution to gain real-time insights into how viewers interact with their platform.

Analyzing customer behavior data — what viewers watch, when, and on which devices — is essential. It enables the delivery of personalized, high-quality viewer experiences that adapt quickly to evolving trends and sudden shifts in user preferences, ensuring one stays ahead in the race to delight audiences. For providers like Go3 leveraging data analytics isn’t just an advantage; it’s necessary in the competitive, ever-evolving streaming landscape.

Go3’s goal was to collect its real-time streaming data across all digital platforms and devices into a centralized system, creating a single source of truth for deep, comprehensive analytics and governance.

Go3 required near-real-time forwarding of specific events to external analytics providers while fully complying with European Economic Area regulations.

The Go3 team faced a formidable challenge: collecting and analyzing over 1 billion monthly events detailing what viewers watch, when, and on which devices. The platform had to handle over 5,000 requests per second from smart TVs, iOS, Android, and web browsers. Flash events like sports matches posed additional difficulties, driving sudden, massive spikes in viewership. Yet, outside these peaks, traffic remained stable, growing during the day and peaking at 8 PM — so designing such a system required scalability, precision, and readiness for unpredictable demand surges.

Solution Building a real-time data analytics platform

When Go3 envisioned turning its ambitious goals into reality, it partnered with AWS expert Chaos Gears. After successfully migrating Go3's services to the AWS cloud, Chaos Gears provided the expertise necessary to design and build a real-time data analytics platform, using AWS’s cutting-edge technologies.

Designing such a system requires deep expertise in processing and analyzing real-time data streams, as well as ensuring scalability and advanced analytics. Peaks in traffic demand precise architectural strategies to manage latency, maintain consistent quality, and support scalability. Understanding diverse data sources, selecting the right technologies, and developing an infrastructure capable of stable yet dynamic scaling was crucial to meeting Go3's high expectations.

The groundwork for actionable data analytics

Our Data & AI team — composed in this case of a Data Architect, a Data Engineer, and a Cloud Engineer — embraced Go3’s challenge with a focus on three core principles: reliability, cost-efficiency, and scalability.

We delivered multiple AWS infrastructure options, ranging from fully serverless to EC2-based solutions, tailored to seamlessly handle Go3’s variable traffic demands.

Guided by these principles and close collaboration with the Go3 team, we developed a real-time data ingestion & integration platform, delivering a robust, scalable, and cost-efficient data ecosystem.

The architecture features a custom-built ingress application deployed on Amazon ECS and EC2 in a high-availability setup. This approach replaces the need for Amazon API Gateway, minimizing costs even under extremely high system throughput. The application efficiently processes all incoming streaming events from Go3’s platforms, validating and pushing them into the data pipeline. The pipeline itself was designed to support both real-time data processing and in-depth historical analysis, ensuring versatile data handling capabilities tailored to Go3's needs.

Behind the scenes of Go3’s data analysis

To address Go3’s needs, we used a suite of AWS services designed to catalog, visualize, and enable seamless data searches while ensuring robust protection against unauthorized access. These tools not only streamline data management but also provide the security and scalability necessary for a reliable and efficient data ecosystem.

We architected a robust solution using Amazon Kinesis Data Streams as the backbone of a whole solution, with Amazon Firehose processing the real-time data. We used a partitioned structure in Amazon S3 for warm and cold storage. We used AWS Glue Crawler to automatically discover and catalog data schemas to ensure efficient data management.

Meanwhile, the AWS Glue Data Catalog maintained metadata for easy discoverability, streamlining management and analysis. For custom transformations, we utilized AWS Glue Jobs/Studio and Amazon Athena to enable SQL-based querying for all historical datasets. Our solution used AWS CloudWatch to create system metrics and visualize them on real-time dashboards, while gathering continuous application logs from the system, ensuring operational visibility. AWS X-Ray provided deep insights into serverless application performance, enabling precise debugging and analysis.

Additionally, Amazon SQS served as a robust messaging queue mechanism, supported by dead-letter queues (DLQs) to handle unprocessed messages. AWS Key Management Service (KMS) was implemented to encrypt data stored securely in Amazon S3, ensuring robust protection against unauthorized access. Amazon DynamoDB, with a single-table design pattern, efficiently tracked the overall status of data ingestion packages.

This architecture empowers Go3 to perform in-depth analysis of historical data while maintaining low query costs through strategic partitioning and compression techniques. To further enhance performance, several AWS Lambda functions consume Amazon Kinesis Data Streams with dedicated enhanced fan-out capabilities to ensure reliable, high-throughput delivery of events to external analytics providers, meeting the demands of large-scale data workflows with precision and efficiency.

A foundation for future needs

The project unfolded in multiple phases, with each stage targeting specific system components and concluding with client reviews and acceptance testing. We delivered the ingress application, the data pipeline, and the components responsible for event delivery to external providers.

Rigorous stress testing was a cornerstone of the process, simulating traffic loads far beyond anticipated peaks to validate the system's reliability and scalability. These tests confirmed the platform’s ability to seamlessly handle traffic surges while scaling down efficiently during low-traffic periods, ensuring optimal performance at all times.

The production deployment was executed in carefully planned stages, starting with the Web, then SmartTV after a few weeks, and gradually extending to iOS, Android, and other platforms.

Also, we focused on creating a system that Go3’s team could seamlessly monitor and manage, without the support of our engineers. To ensure seamless system oversight, we implemented a comprehensive monitoring solution through a single dashboard. This provides real-time insights into system health and performance metrics, empowering Go3’s team to track usage patterns such as the number of viewers and their preferences. It also allows detailed visibility into the types of events being processed, bridging operational efficiency with business intelligence.

The architecture was designed with adaptability at its core, enabling Go3 to introduce new data points or modify existing ones seamlessly — no code changes, no downtime. Designed for growth, the system is future-ready, with the capacity to accommodate additional data consumers as Go3's needs evolve.

Outcome The right content, at the right time

Go3 understands that the success of an OTT service depends on delivering content that resonates with its audience. Using data-driven insights, the company has built a recommendation system and marketing automation platform, ensuring personalized viewer experiences that drive engagement and satisfaction. Making informed decisions based on data enhances the viewer experience, drives operational efficiency, reduces costs, and solidifies a competitive edge of Go3 in the market.

The platform delivered by Chaos Gears met the client’s expectations, transforming Go3's ability to analyze and respond to viewers' behavior. The solution provides real-time insights into engagement across all platforms, SQL-based access for in-depth historical analytics, and automated schema detection with catalog updates. Our custom solution handles their scale needs and throughput, it integrates seamlessly with external analytics providers and adheres to strict European data regulations. Its modular, scalable architecture ensures that Go3 can evolve its analytics capabilities without significant architectural changes or cost increases, empowering its data-driven growth.

As a major player in the world of streaming, with over 600k subscribers at the beginning of 2025, Go3 is solidifying its position as the leader in the OTT market in the Baltics. The Go3 team understands that the real power of data comes from how effectively it is communicated and implemented to enhance user experiences.

So, don’t be surprised if the perfect show or movie from Go3’s lineup keeps you glued to your screen for just one more episode — it’s all by design ;)