Ideaticon / Case Study / Mythical Games

100 million daily events, with the game still running.

Mythical Games needed one behavioral and economy data system across four product teams. I built the telemetry and governed transformation layer around Snowplow, GCP, CockroachDB and dbt. It processed more than 100 million events a day at 99.99% uptime and supported realtime economy monitoring, analytics, and downstream ML features.

99.99% Uptime
100M+ Daily Events
4 Product Teams

The volume was only one problem.

Ingestion was only the first problem. Wallet activity, marketplace transactions, player behavior and game economy events had to arrive quickly, keep their meaning, and resolve to definitions that four product teams could share. I owned the platform work from collection and storage through transformation, quality controls, and data prepared for ML.

Events designed for game economies

The Snowplow implementation captured NFT transactions, wallet interactions, marketplace activity and player behavior with blockchain enrichments. A shared schema registry made the contract explicit and stopped each team inventing a different meaning for the same event.

100M+ events a day on GCP

Dataflow and Pub/Sub handled the realtime flow, including peaks around live game activity, while CockroachDB provided the distributed operational store. The production system sustained more than 100 million daily events with 99.99% uptime.

Shared definitions across four teams

A common dbt framework governed more than 200 models, with ownership, contracts, and quality gates attached to the work. Player value, economy balance and fraud signals came from one transformation layer rather than four parallel interpretations.

Operations built for live games

Storage across multiple regions, automated recovery and observable pipelines kept routine failures away from the product teams. Scaling and failover were built into the platform, not left for release night procedures.

Data built for specific ML uses

The platform produced governed features for fraud detection, churn prediction and recommendations. Teams had versioned, reproducible inputs they could train and evaluate against.

Business impact

  • More than 100 million behavioral and economy events processed each day
  • 99.99% uptime for the production data system
  • Four product teams working from shared schemas, models, and ownership rules
  • Realtime economy and fraud monitoring for game marketplaces
  • More than 200 governed dbt models behind analytics and ML features