2M+
meter readings processed daily
99.95%
pipeline reliability
-35%
billing-related support calls
~40 min
sooner usage anomalies surfaced
The Client
A regional utility serving hundreds of thousands of residential and commercial customers, midway through a multi-year smart-meter rollout. The new meters reported consumption in short intervals around the clock — replacing monthly manual reads with a continuous stream of data the utility's existing systems were never designed to handle.
The Challenge
The rollout was succeeding as a hardware project and failing as a data project. Interval readings landed in the meter vendor's head-end system and mostly stayed there; billing still ran on periodic extracts, and the operations team had no view into the stream at all. When a meter misreported or a customer's usage spiked abnormally, the first anyone heard of it was an angry phone call after the bill went out — and billing disputes were climbing as more smart meters came online.
Customers, meanwhile, could see none of the data their own meters produced. The utility's IT department was capable but fully committed to core systems, and had no in-house experience with streaming data platforms. They needed a partner to design and build the platform, then hand it over cleanly to their team.
What We Did
Dillo delivered the platform as an outsourced data-platform project — data engineers, a backend engineer and a delivery lead, working to milestones agreed with the utility's IT director, with knowledge transfer to the internal team built into the plan from the start.
- Built streaming ingestion on Kafka from the meter head-end system: validated, deduplicated interval readings flowing continuously, with dead-letter handling and replay so no reading is silently lost.
- Implemented Spark processing into BigQuery, with dbt models producing certified consumption datasets for billing, operations and analytics — one governed source of truth instead of ad hoc extracts.
- Added billing-anomaly detection: statistical checks against each meter's own history and peer profiles flag misreads, stuck meters and abnormal spikes for review before invoices are generated, not after complaints arrive.
- Launched a customer usage portal in React, giving customers daily and hourly consumption views, bill projections for the current cycle, and high-usage alerts they can opt into.
- Hardened the platform for utility operations — monitoring, alerting, SLA dashboards and runbooks — and ran paired handover sessions so the utility's IT team operates it day to day.
The Results
- The platform processes 2M+ meter readings daily, scaling smoothly as the rollout adds meters each month.
- 99.95% pipeline reliability, with replay and dead-letter handling ensuring readings are recovered rather than lost when upstream systems hiccup.
- 35% fewer billing-related support calls: anomalies are corrected before bills go out, and customers can answer "why is my bill high?" themselves in the portal.
- Usage anomalies now surface about 40 minutes sooner on average, letting operations catch stuck meters and abnormal spikes while they are still cheap to fix.
- The utility's own IT team now runs the platform, with Dillo on a light advisory retainer — the handover the client asked for on day one.
Figures reflect outcomes reported for this engagement; they are project results, not audited benchmarks.
Tech Stack
Services Used
Software Outsourcing — a data-platform delivery team building the system end to end, with structured handover to the client's IT organization, on transparent pricing.