How a Global Manufacturer Cut Operational Decision Time by 28%
A global manufacturer of industrial equipment was running plants it could not see into. Production data, equipment telemetry and supply chain metrics sat across ERP systems, IoT platforms and spreadsheets, so plant managers made decisions from reports that described a situation that had already moved on. OnPoint Insights consolidated it into a Microsoft Fabric Lakehouse on OneLake, with real-time ingestion from the equipment and Power BI dashboards on top.

Plant performance, live
IoT telemetry and production data stream into the platform, so throughput and equipment utilization are visible as they happen rather than in the following day’s report.
ERP, sensors and spreadsheets in one lakehouse
Production, supply chain and equipment data centralised in OneLake, so plants can be compared instead of collated.
28% faster operational decisions
Decisions that used to wait on manual data preparation now happen more than a quarter sooner.
The problem
Decisions made from yesterday’s report
A plant manager’s decisions are perishable. A bottleneck that formed this morning needs addressing this morning, and a report describing it tomorrow is a record rather than a tool. The data needed to spot it almost always exists already, scattered across the systems that produced it.
Our client had that data across ERP systems, plant sensors, operational databases and spreadsheets, spread over multiple facilities. The legacy analytics platform could not process streaming data at all, so manufacturing leaders were working from delayed reports and manual preparation, with no way to see where a plant was losing time.
What was getting in the way
- Production data existed across multiple disconnected systems, including ERP, plant sensors and operational databases.
- The existing analytics platform could not process streaming data from equipment and manufacturing systems.
- Manual data preparation and fragmented reporting slowed operational decision-making.
- Without centralised analytics, identifying inefficiencies and bottlenecks across plants was close to impossible.
Our approach
One platform for the stream and the history
What we found
The legacy platform could not process streaming data, which meant no amount of work at the reporting layer would have produced live insight. The platform had to change before the dashboards could be worth building.
What we did
Telemetry straight into the platform. Real-time ingestion pipelines bring IoT telemetry and production data into Fabric as the equipment generates it.
One lakehouse across the network. Production, supply chain and equipment data centralised in a Fabric Lakehouse on OneLake, so performance at one plant can be set against another.
Dashboards on the metrics that run a plant. Power BI tracking plant performance, production throughput and equipment utilisation, for the people who act on them rather than for analysts.
Anomaly detection and access control. Predictive analytics to flag production anomalies and potential equipment failures, with governance policies and role-based access over operational data.
The choice we made
We built the predictive layer on the same lakehouse that feeds the dashboards, rather than in a separate data science environment beside it. One copy of the data means the anomaly model and the plant manager’s dashboard cannot disagree about what a machine did, and the model’s inputs stay current without a second set of pipelines to maintain. Two environments is the more familiar pattern, and it is how prediction and reporting end up telling different stories.
-
Fabric Data Factory -
Microsoft Fabric -
OneLake -
Power BI -
Azure IoT Hub -
Fabric Real-Time Intelligence
The results
What changed across the plants
Plant performance became something managers could act on during a shift rather than review after one.
Every plant visible at once
Operations leaders see production performance across facilities in one place, which makes comparing plants a question rather than a project.
Fewer disruptions
Monitoring equipment as it runs allows issues to be caught early, which reduced production disruptions across the network.
Decisions in the moment
Plant managers address performance issues as they appear instead of waiting for a reporting cycle, and the time taken to reach an operational decision fell 28%.
Facing a similar challenge?
Talk to our data and AI team about where you are today and what good looks like. 30 minutes, no obligation.
Book a consultation
