Data pipelineIn progress
Accounting data, to the minute
Entries from the accounting software reach a PostgreSQL warehouse in under a minute, then feed one Power BI model shared by every client. I automated how the data is pulled in, built and served; the Data team owns the processing.
- Role
- Pipeline automation
- Client
- Accounting firm, Paris
- Year
- 2026
- from source to warehouse
- < 1 min
- to read a month of entries (58,757 lines)
- 0.9 s
Context
Reports were refreshed from files synced through an office PC with a slow connection.
What I built
- 01Incremental sync from the accounting API's change logs
- 02Automated warehouse build, from raw data to reporting tables
- 03Delivery to Power BI, rewriting only the periods that changed
Key decisions
Measure before choosing
Real data showed 7% of corrections landing on months older than the last two, so a simple rolling refresh would have missed them without anyone noticing.
Outcome
Rolling out, one client group first.