Big Data Analytics for Business Intelligence
Predictive dashboards that shortened a reporting cycle from 3 weeks to 2 days.
Key Details
| Challenge | An industrial supplier waited three weeks for a reporting pack assembled by hand. |
|---|---|
| Solution | A modeled warehouse and Power BI layer the ops team could refresh themselves. |
| Technologies | Azure Data Factory, Synapse, Power BI |
Technologies used
Azure
Power BI
Python
SQL
dbt
Client background
An Australian industrial supply business assembled a three-week reporting pack by hand across ERP extracts and Excel. Leadership decisions lagged the market, and every new KPI meant another fragile spreadsheet.
Key challenges
- Reporting depended on a few analysts who knew undocumented Excel logic.
- ERP extracts were inconsistent; numbers rarely matched between packs.
- Ops could not self-serve — every question queued behind the reporting cycle.
- No semantic model, so “revenue” meant different things to different teams.
What we built
- Azure Data Factory pipelines into a Synapse warehouse with tested transforms.
- A single semantic model for sales, inventory and margin.
- Power BI dashboards ops could refresh without analyst intervention.
- Documentation and ownership so new metrics did not fork the model.
Project team: 7 engineers across AI/ML, backend and domain specialists — delivery over 20 weeks.
How we delivered
01
Audit
Inventory of reports, owners and the decisions each pack actually drove.
02
Warehouse
Modeled facts and dimensions that matched how the business operates.
03
BI layer
Shipped Power BI with certified datasets and row-level access.
04
Handover
Trained ops to refresh and request changes through a governed path.
Business impact
3wk→2dReporting cycle
OpsOwns refresh
OneSemantic model