IoT

Connecting an industrial equipment fleet and making it predictive

The context

A French industrial equipment manufacturer builds equipment installed at end customers across the country. Until now, a failure meant a phone call, a trip, an on-site diagnosis, and sometimes a second visit because the right part was not in the van.

The manufacturer wanted two things: to know remotely how its equipment actually behaves in service, and to be able to hand that data back to its customers, who are starting to ask for it and whom European data regulation now entitles to it.

What we did

We added cellular telemetry from the equipment into AWS IoT Core, then built the platform that makes use of it.

Equipment telemetry ingested continuously, aggregated in batch, served three waysDeployed equipment streams telemetry into IoT Core, from where Kinesis Firehose lands it on S3 continuously. A batch stage consolidates those measurements. The consolidated data is then served through three channels from the same source: a documented REST API for end customers, an MCP server for agents, and the web interfaces used for fleet management and after-sales.From equipment telemetry to the interfacesEquipment AtelemetryEquipment Btelemetrydeployed fleetContinuous ingestion — IoT Core, Kinesis Firehose, S3Batch aggregation — consolidated measurementsREST APIend customersMCP serveragentsWeb interfacesfleet and after-sales
Ingestion is continuous, aggregation is batch: measurements arrive one by one but are read a fleet at a time. Splitting the two is what lets the read side be reworked without touching the pipe — and the three channels are three views of one dataset, not three pipelines.
  • Ingestion and storage of the measurements, with a data model designed so that adding a new equipment type does not mean rebuilding the pipeline.
  • Predictive maintenance trained on operating history, so an intervention is triggered before the failure rather than after the customer’s call.
  • Documented data portability APIs for end customers, built to satisfy the obligations introduced by the Data Act.
  • Three read channels over one source: those REST APIs for end customers, an MCP server for agents, and the web interfaces used internally. Three views of one dataset, not three pipelines to keep in step with each other.
  • Administration interfaces for fleet management and after-sales tracking.
From consolidated history to an intervention before the failureThe consolidated measurement history held on S3 feeds a SageMaker model, which detects the trends that precede a failure. The resulting probabilities are shown in the web interface the teams already use, and a technician acts on them before the equipment fails rather than after the customer calls.From history to the interventionHistory on S3consolidated dataSageMaker modeltrends detectedWeb interfacefailure probabilitiesTechnicianintervenes in time
The chain is only worth its last step: a probability that does not reach a technician in time changes nothing. Which is why the alert threshold is a business setting, not a model parameter — it trades false alarms against missed failures, and only the maintenance team can price that trade.

The outcome

Interventions are now planned rather than endured, and the support desk answers with the full history of the equipment in front of it.

Data portability, approached at first as a regulatory constraint, turned out to be a commercial argument: the manufacturer now sells visibility over its equipment as a service, to customers who needed it without knowing how to ask.

The delicate part was not the predictive model but the alert threshold. It trades false alarms against missed failures, and that trade belongs to the maintenance team rather than to the model: a business setting, exposed as one, rather than a parameter buried in a notebook.

All case studies

Talk to us directly

Write to us directly. No salesperson, no qualification call. One of us two replies, within 24 hours.

contact@onescale.io

We reply within 24 hours, and it is one of us who replies.

Based in Lyon, working with clients in France, across Europe and internationally.