Everyone has data. Not everyone trusts it. RBC Data Platform is an enterprise data warehouse on open technologies from RBC Group. It comes with RBC Data Portal: one window that shows where every number came from, who owns it and why it can be trusted.
In brief
- The warehouse is built in layers from source to data mart: Stage, ODS, Raw Vault, Business Vault and marts. A new source is added without reworking what is already in place.
- The portal gives a single entry point to DWH administrators, data owners, business users and the acceptance committee: navigator, metadata catalogue, quality quarantine, monitoring, acceptance testing.
- For reporting and dashboards we connect Apache Superset, an open BI platform with no per-user licence fees.
What the solution includes
- An enterprise data warehouse built on Data Vault 2.0, with calculation rules as code and automated tests.
- A management portal: one entry point to the warehouse for every role.
- Quality control with quarantine and an owner for every data route.
- Monitoring of loads, quality and infrastructure with ready-made alerts.
- Analytics: Apache Superset dashboards on top of the marts.
Architecture: from source to data mart
Data follows a clear path: raw copies of the sources (Stage), a cleansed operational layer (ODS), a historised core (Raw Vault and Business Vault) and marts for reporting. Reference data and attributes maintained by the business are kept separately.
The technology foundation is open: StarRocks and Apache Iceberg for storage and queries, dbt for calculation rules and tests, Airflow and NiFi for loads. Calculation rules are written as code, so every change is visible, testable and reversible.
Warehouse layers and what each role sees through the portal
The portal: one window into the warehouse
- Sign-in and permissions. Local accounts, Active Directory or OIDC; roles come from groups. Permissions are checked on the server for every request, personal data is masked, and every action goes to the audit log.
- Layer navigator. Each object card shows columns, data, model code and lineage.
- SQL console. Read access across all layers; write access only for a dedicated role, with confirmation and audit.
- Metadata catalogue. A business glossary with four-eyes approval, a report catalogue, lineage down to the field and change impact analysis.
- Marts and reports. Filtered views, export to Excel, regulatory reports by reporting date with date-to-date and period-to-period comparison.
- Reference data without a developer. Business users maintain reference data in an Excel-like grid or upload files; every load goes through validation and quality control.
Quality quarantine: bad data does not slip through
If checks hold rows back on their way into the warehouse, the data owners are notified. They respond “taken into work” or “fixed at source”, and the load is released with a second signature. Owners are assigned per route, so the question “whose data is this” never hangs in the air.
Monitoring: what is happening to the data right now
Monitoring dashboards open right inside the portal: loads, quality and deadlines, the state of the database, the orchestrator and the servers. The delivery includes ready-made alerts for load completeness, pipeline failures and connector health, plus a runbook that says what to do for each. Configurations are prepared for development, test and production environments.
Analytics on top of the warehouse
The marts become the basis for dashboards in Apache Superset, an open BI platform with no per-user licence fees.
The metadata catalogue and glossary give every metric an agreed definition, an owner and a lineage. Such data is also fit for AI tools if the customer decides to connect them inside its own perimeter.
Acceptance without surprises
Acceptance testing is built into the portal: test cases with role-based instructions, control queries that run at the press of a button and get a status at once, protocols in Word and Excel. The test lead sees a summary: what has passed, what is in progress, what is overdue against the plan.
Who it is for
- Banks. Regulatory reporting and data marts, including the requirements we covered in the article “ARRFR data marts”.
- Telecom companies. Billing, subscriber base, network and finance data in one warehouse: the cost of each service, payback on equipment and network investments, and tariff justification for the regulator.
- Holdings and groups. Unified reporting from different accounting systems.
- Manufacturing and distribution. Data from ERP, planning and HR systems in one place: Oracle JD Edwards, IBM Planning Analytics, RBC HCM Solutions.
Frequently asked questions
Will we be locked in to the supplier?
The solution is built on open technologies and open data formats.
Can the warehouse run in a closed environment?
Yes. The solution is deployed in the customer’s infrastructure, with separate development, test and production environments.
How does a project start?
With a source assessment and a first scope: a few systems, one or two marts, a working portal. From there the warehouse grows source by source.
See the portal live
RBC Group builds enterprise data warehouses for banks, telecom companies and enterprises in Kazakhstan, Uzbekistan, Ukraine and Azerbaijan. We will demonstrate the portal and discuss the first scope for your data.
Contact usmarketing@rbc-grp.solutions, +7 727 258 1712
Technology foundation
- StarRocks — the analytical MPP engine: reports and data marts are calculated quickly even on large volumes.
- Apache Iceberg — an open storage format: the data stays readable by other tools and is not locked inside one system.
- dbt — calculation rules are written as code with tests: every change is visible and verifiable.
- Apache Airflow and Apache NiFi — scheduled, controlled loads: you can see what was loaded, when and with what result.
- OpenMetadata — the data catalogue: every metric has a description, an owner and a lineage.
- Apache Superset — dashboards and reports for the business with no per-user licence fees.