Skip to content

Grid analytics over 2.6 billion readings a day

Python, Kafka and ClickHouse for a Dutch grid operator

Voltgrid · Netherlands
Client
Voltgrid
Industry
Energy & Industrial IoT
Year
2026
Duration
10 months
Team
8
Engagement
Time & material

Results

2.6B

Readings per day

4 s

Anomaly alert latency

-68%

Storage cost per billion rows

340 ms

p95 dashboard query

The challenge

Voltgrid's analytics ran on nightly batch jobs. Operators saw grid anomalies the next morning, which is too late to act on a load imbalance, and storage cost grew 40% year over year.

What we built

Kafka ingest with schema registry, stream processing in Python, and ClickHouse for storage with tiered retention. Anomaly detection runs on rolling windows and pushes alerts to the control room in under four seconds. We rewrote the storage layout to cut cost per billion rows by 68%.

The outcome

Operators now see anomalies within four seconds instead of the next day. Storage cost fell 68% per billion rows despite ingest volume tripling.

“They tripled our ingest volume and cut our storage bill by two thirds in the same quarter. The cost review they run every month has become a standing item in our leadership meeting.”
SD

Sanne de Vries

Head of Data Platform, Voltgrid

DirectNetherlands

Next case study

Low-latency trading terminal

→

Tell us what you are building

Send a brief and a solution architect replies within one business day — with questions, not with a template.