When your database handles the load but blocks your decisions

Blank is growing. The Crédit Agricole group neobank for freelancers (business account, automated VAT, built-in URSSAF (French social security) contributions) onboards new customers every week. But internally, the teams are operating in the dark: no reliable dashboards, CSV exports done by hand, decisions made on instinct.

The problem is not a lack of data. It is an architecture designed to do something else.

The DynamoDB dead end: the core database is an operational success: low latency, scalability, availability. But DynamoDB is not built for analytical aggregations. A table scan in production means latency injected into the customers’ critical path. Non-IT teams cannot write ad hoc queries. Every analysis request goes up to the developers, and slows everyone down.

The solution: separate the planes. Nobori Partners designs a decoupled architecture: DynamoDB Streams captures every write in real time without touching the source table. Lambda functions transform and route the events to dedicated analytical storage. Metabase exposes a SQL interface accessible to operations staff, with no ticket and no waiting.

From POC to production in 4 months: the solution is fully declared as Infrastructure as Code (CloudFormation). Reproducible, versionable, auditable. Knowledge transfer is built into the engagement: the internal teams take over from go-live.

The result: 1 FTE freed up on back-office processing, business teams self-sufficient with their data, and an analytics architecture that evolves on its own, without ever interfering with production.