Cloud Imperium Games Levels Up Data-Driven Decision-Making with ChaosSearch

Cloud Imperium Games is an international, independent game company. The company is currently developing Star Citizen, a record-shattering crowd-funded title that combines classic space sim gameplay with amazing visuals; and Squadron 42, a Hollywood-caliber single-player game set in Star Citizen’s universe.

Cloud Imperium’s analytics team is on a mission to expand data literacy and bring data-driven decision making to every aspect of development — from the end game economy and trading, to the weapon balance and fight mechanics, to the new player experience. Find out how they did it by transforming their Amazon Simple Storage Service (Amazon S3) cloud object storage into an ultra-performant app analytics platform with ChaosSearch.

The Challenge

Crafting new universes is no small feat. Harnessing the potential of game data is essential for achieving success. “We’re building a monumental project,” said Keegan Standifer, data solutions manager at Cloud Imperium Games. “Good data is everything. The more data that is available, the better of a game we can build for our players.”

The team knew that leveraging application telemetry data for analysis would be critical to not only fixing bugs and performance issues, but also understanding in-game player behavior. Using this analysis, they could make changes to the game design and mechanics to make play more balanced and challenging. Ongoing changes and improvements would bring new players to the game, and keep loyal players coming back for more.

“Our community lets us know about what they enjoy about the game, but they're also not bashful about making sure that we have visibility into issues we need to improve,” said Standifer. “It's our job to put the perceptions of the players into a quantifiable data format that shows decision-makers which features or improvements we need to be focused on.”

However, the team had access to so much data that their current self-service analytics system became untenable. As a result of the game's popularity, the metrics, events, and logs generated to track every single action and aspect of gameplay suffered an explosive growth in volume and diversity. “Prior to ChaosSearch, there were significant restrictions on how much data we could keep hot in our self-service solution,” said Standifer. “We couldn’t handle many requests because of the short retention of our hot data availability. When those requests would come to the analytics team, we would have to build custom or manual solutions. For example, we would build ETL processes that extracted the desired data and made it available in the format that our different developers wanted.”

The sheer volume of data put a strain on the incumbent log analytics solution and forced the team to make difficult trade-offs around data retention and flexibility. The cluster’s performance and stability was faltering. Facing these challenges, the team decided to evaluate other solutions to not only remove data retention limits, but also add new capabilities and flexibility.

The Solution

During proof of concept tests, Cloud Imperium evaluated the ChaosSearch platform side-by-side against their current platform and other competitive solutions. Using ChaosSearch, the team experienced the freedom and power to analyze large amounts of data, without the constraints of retention limits. ChaosSearch not only met the team’s performance goals, but also had almost no learning curve for the analysts. Combined with the economic advantage of Amazon S3, the team found ChaosSearch the clear winner.

ChaosSearch works by transforming the company’s existing cloud object storage in Amazon S3 into a hot data lake, with game analytics readily available for developers. This avoids the painful process of data movement, which put major stress on the data analytics team in the past. Now engineers can truly self-serve, using tools of their choice to analyze the data they need, when they need it — at scale.

“The more accessible your data is, the more likely people are going to naturally integrate it into their workflows,” said Standifer. “Going to a self-service portal like ChaosSearch lets developers explore data with an intuitive user interface where they can pick out the information they want, build visuals quickly, see the events and analytics they're interested in, and dig into it field by field. When you're releasing new versions of a game or balancing a feature, you want to see how it impacts the player experience before or after.”

In addition, the ChaosSearch team was able to easily migrate more than 700 analytics artifacts (charts and dashboards) from the previous solution, minimizing any potential friction during the transition.

The Impact

Today, ChaosSearch plugs directly into Cloud Imperium’s Amazon S3 buckets, allowing access to the entire dataset of logs and events. This game analytics solution enables the team to analyze hundreds of terabytes, or several years’ worth of platform and player interactions, including application, infrastructure, and player event data. Analysis occurs at a fraction of the cost of CIG's previous solution, with much less technical management and oversight required for the platform itself.

Developers now have access to vast amounts of data via Search (Elastic API), SQL (Presto/Trino API) and Conversational AI (GenAI API), as well as embedded OpenSearch Dashboards (Elastic) and Superset (SQL BI) tooling. “With ChaosSearch, we can preserve the interfaces and the tools that our developers enjoy using,” said Standifer. “That way, developers can explore the data that we do have, and the analytics team can focus on the more advanced questions and complex data problems that require an experienced analyst.”

Analyzing this data greatly enhances the team’s ability to make ongoing development and commercial decisions. “We’re able to quantify player sentiment over different periods of time to understand if feedback is getting more positive or negative after certain improvements,” said Standifer. “There's a lot of different ways to digest this information, and we're fortunate that we can start to combine different dimensions of our data for use in our development process.”