Firebase Analytics is a fast way to start tracking events, but many teams eventually hit limits in cost predictability, raw data control, and long-term portability.
The practical alternative is not a giant internal platform from day one. A smaller path works better: keep ingestion generic, land raw events in durable storage, and keep your downstream query layer flexible.
In Rawbbit, events come in through a simple HTTP endpoint, get buffered with NATS JetStream, and are written as partitioned Parquet files in object storage. That raw layer becomes the stable source of truth.
From there, you can query using BigQuery external tables and add modeled tables gradually. This gives teams room to evolve analytics without locking event history into one vendor contract.
The key idea is ownership: own the pipeline runtime, own the raw files, and keep transformation logic explicit. You can still move fast, but with lower long-term migration risk.