Keeping costs under control on a growing product is rarely about one big optimization.
In practice, the useful work is usually smaller and less glamorous:
- reducing duplicated event flows
- checking which queries deserve to exist
- revisiting data retention assumptions
- questioning whether every dashboard really needs fresh data
On a retail SaaS stack built around Firebase and BigQuery, those questions matter because convenience can become expensive very quietly.
The lesson is not “never use managed analytics”. The lesson is to treat data architecture as a product decision, not just a backend detail.