When Every Shot is a Data Point
Inside Embark Studios' data platform, and how gaming scale forces you to get the fundamentals right
Stockholm sent off the season with a surprise rain shower right as people were making their way to Embark Studios’ office. Some arrived drier than others. Either way, the room filled up fast, drinks were flowing, food was out, and the energy was good from the start. New faces, familiar ones, and at least one person who made the trip all the way from Linköping.
Before we got into the talk, we shared some news the community has been waiting for.
The Data & AI Stockholm Summit is happening on October 14th at Space Stockholm. A full day, practitioner-first, no sales pitches. Everything the community stands for, just bigger. Read more about it on the Official Summit Page.
A game studio and its data problem
Simon Carlson, Software Engineer at Embark Studios, gave the talk for the evening. Embark makes The Finals and ARC Raiders, two multiplayer games that between them generate a volume of data that reframes what “a lot of data” actually means.
We track every shot.
Not figuratively. Every weapon fired, every player position, every in-game interaction, across all active sessions, continuously. The numbers Simon shared set the scene quickly: over a thousand distinct event types, roughly 100 billion events landing every single day, and all of it needs to be queryable in under two seconds.
The reason this matters goes beyond engineering pride. Game designers at Embark need to know whether a weapon is overpowered. Map designers need to see where players are actually going versus where they were supposed to go. Security teams need to catch cheating. None of those questions get answered well without the data infrastructure to support them, and none of that infrastructure is optional.
How they actually built it
Simon walked through how Embark gets from a raw in-game event to something a designer or analyst can act on, and the interesting part was less the scale and more the restraint.
The previous ingestion system was expensive, slow to debug, and leaned heavily on managed services the team had limited control over. The replacement was simpler on purpose. Events leave the game server, pass through a lightweight sidecar process running alongside it, hit Pub/Sub, and land in BigQuery. Fast, controllable, and stable since day one of the new system going live.
On the other end, the team built their own dashboard infrastructure on top of DuckDB, which Simon was fairly enthusiastic about. The tooling that sits on top ranges from live dashboards to full map replays showing exactly how a round played out, to heatmaps that reveal player movement patterns across an entire map. Different people at Embark consume the data differently, and the platform is built to serve all of them.
Simon wrapped that section with one of the cleaner takeaways of the evening:
Code > UIs.
Defining your cubes, your metrics, your pipeline logic in code rather than through graphical interfaces keeps things version controlled, reviewable, and reproducible. It is an obvious point once someone says it out loud, but it is also one a lot of teams learn the hard way.
The takeaway that landed
Simon closed with something that resonated with a room full of people who have all, at some point, stared at a slide with hundreds of data tool logos on it wondering where to start.
Your data stack does not need to look like that.
Pick solid foundational technologies. Build for your actual needs. Accept that sometimes the right move is to build something yourself rather than add another tool. The complexity creep in data platforms is real, and Embark’s approach was a useful counterpoint to it.
The room takes over
After Q&A the evening moved into group discussions, which by now is a signature part of how Data & AI Stockholm runs.
Three questions anchored the conversations
What architectural decision had the biggest impact on your own data platform journey? What does an AI-ready semantic layer actually mean in practice? And has the rise of AI genuinely changed how you think about data architecture, or are the fundamentals still the same?
The conversations that followed were some of the most candid of the season. People brought real examples, real mistakes, and real opinions.
Each group shared back to the room before the evening wound down with a last round of drinks and a natural drift into smaller conversations that nobody seemed in a hurry to end.
And that is a wrap on the season
This was our last event before summer, and it was a good one to go out on.
While you are enjoying the break, it is a good time to catch up on the Community Stories podcast and articles that have been building up on the Data & AI Stockholm channels. Some genuinely great conversations and reads in there if you have not had a chance to go through them yet.
We will be back in August, and behind the scenes we are already deep into building what October 14th is going to look like. The Data & AI Stockholm Summit at Space Stockholm is shaping up to be something the community will remember. To stay up to date with the agenda, speakers, and everything else, check it all on the official page.
Thank you to everyone who has shown up this season, asked the hard questions, stayed for the conversations after, and made every one of these evenings worth having. This community is what it is because of the people in the room.
See you on the other side of summer, and see you on October 14th.
Data & AI Stockholm 💙
Stay Connected
If you’d like to stay connected with the community and hear about future events, the best places to find us are: