Refining raw OCPP data streams into decision-ready dashboards

As electric vehicle charging infrastructure scales, the volume of raw data generated by charging points has increased significantly. Metergram developed a modern data platform to transform this complex stream of information into actionable insights. By using Microsoft Fabric and a Medallion Architecture pipeline, we built a system that organizes raw records to optimize operations and predict equipment failures. This approach allows charging providers to turn technical data into a reliable tool for strategic decision-making.

Challenge

Each time a charger starts a session, reports a fault, or plugs in a connector, it creates data. Multiply that by thousands of chargers running 24 / 7 and you get a massive firehose of information originating from:

real-time OCPP events via Kafka
REST APIs that supply structured yet isolated metadata
semi-structured payloads such as JSON and CSV

The volume was exciting, but the complexity was daunting. We had to answer three questions:

How do we capture and organize all of this data in one place?
How do we clean it, enrich it, and add context?
Most importantly, how do we make it valuable to the business?

Approach

We avoided a patchwork of tools that would need rework later. Instead, we chose Microsoft Fabric, an integrated cloud platform for data engineering, storage, and visualization. It handles both structured and semi-structured data in a single, scalable ecosystem.

To guide the data flow we used Medallion Architecture.

Bronze Layer: A landing zone for raw, untouched data. Here, nothing is thrown away, just in case we need to revisit the source.

Silver Layer: We cleaned it, filtered it, and picked out what really mattered like session IDs, connector statuses, andtimestamps.

Gold Layer: This is where the magic happens. We created well-modeled tables that gave context to everything – using slowlychanging dimensions (Type 2) to preserve history and power analytics.

Solution

Data Ingestion: Kafka events and API pulls feed directly into the lakehouse.

Built-in ETL: Fabric notebooks and pipelines transform raw payloads into tables ready for Power BI.

Analytics Layer: Power BI connects to the Gold tables so users can monitor charger health, usage, and energy consumption without writing SQL.

The entire organization now shares a single, trusted source of truth that supports timely and informed decisions.

Layer	Purpose	Key Actions
Bronze	Raw landing zone	Store untouched data for full auditability
Silver	Clean and refined	Filter, deduplicate, and add essentials such as session IDs and timestamps
Gold	Analytics ready	Model facts and Type 2 dimensions to support BI and machine learning

Outcome

Faster Decisions:
Accurate dashboards let stakeholders act rather than react.

Automated Workflows:
Scheduled ETL keeps data fresh without manual effort.

Ready to Scale:
Whether we add 10 or 10 000 chargers, the platform grows effortlessly.

What’s Next

We plan to integrate AI models for anomaly detection and predictive maintenance, add real-time alerts, and develop advanced forecasting. These enhancements will further strengthen a smarter, cleaner, and more connected EV ecosystem.