Your machines are already talking
Vibration, temperature, power draw, work cycles: every line produces data that almost always goes unused. We collect it and turn it into real-time alerts on out-of-norm behaviour, before it becomes a machine stop — and, where history allows, into failure prediction.
We connect production lines and machinery to turn the signals they already produce — vibration, temperature, power draw, work cycles, downtime — into useful alerts. We start with real-time anomaly detection, which needs no historical data and flags out-of-norm behaviour before it becomes a machine stop; in parallel we build the dataset that makes predictive maintenance possible. Retrofit on existing plants: IIoT sensors where needed, direct reads from PLCs and SCADA where the data is already there.
- ✓You can start without history: anomaly detection learns the machine's normal behaviour and flags drift right away
- ✓The dataset grows while the system is already running: once enough failures are collected, predictive models plug in without rebuilding the setup
- ✓No machine replacement: we read what the plant already exposes and add sensors only where data is missing
- ✓Production data can stay on site: on-premise or edge processing, cloud is a choice and not a requirement
Type, timeline, pricing and stack
| Project type | Real-time anomaly detection, predictive maintenance, shop-floor dashboards, downtime alerts |
|---|---|
| Typical timeline | Pilot on one line in 4-8 weeks, progressive rollout per department |
| Price range | Pilot on one line from €8k · Plant rollout on request |
| Typical stack | IIoT sensors, MQTT, Modbus, OPC-UA, TimescaleDB, InfluxDB, Grafana, Python (scikit-learn, PyOD), edge computing |
Frequently asked questions
Do we need failure history to get started?+
No. If the history exists, we go straight to predictive models. If it does not — the most common case — we still start with live anomaly detection, which learns the machine's normal behaviour and flags deviations without any past data. Meanwhile the system builds the history that will make failure prediction possible.
Do we have to replace machinery or the PLC?+
No. We work as a retrofit on existing plants: where the PLC or SCADA already exposes data we read it directly (Modbus, OPC-UA), and where data is missing we add non-invasive IIoT sensors such as accelerometers, temperature probes and current clamps. Installation is planned around downtime you have already scheduled.
What does a monitoring setup cost?+
A pilot on one line — sensors, gateway, data collection, dashboards and live anomaly detection — starts at €8k. Full plant rollout depends on machine count, data already available and network infrastructure: we quote it after a site visit, not from a price list. That is part of what the pilot is for: measuring the value before extending.
How long before we see something useful?+
The first dashboards with real data from your line arrive within a few weeks of installation. Anomaly detection needs an observation period to learn normal operation — typically a few weeks of ordinary production — after which it starts flagging. Predictive failure models take longer, because they need observed failure events.
Does our production data leave the company?+
Only if you decide so. The default architecture processes data on site with a gateway in the plant: only what is needed for remote dashboards goes to the cloud, and only if you want it to. For strict constraints the whole setup can stay on-premise. We put the data flow in writing before starting.
What is the difference between anomaly detection and predictive maintenance?+
Anomaly detection looks at now: it says the machine is behaving differently than usual, without knowing why. It needs no history and works immediately. Predictive maintenance looks ahead: it estimates when a component will fail, and to do that it must have seen past failures. They are two steps on the same path, not alternatives.
What if the system produces false alarms?+
That is the main risk with these systems: one that cries wolf gets ignored, and at that point it is worse than not having it. This is why sensitivity thresholds are tuned together with your maintenance team over the first weeks, starting conservative, and every alert is checked against the people on the machine. Tuning is part of the project, not an extra.
How we work
Our agency process in 5 steps
Discovery & Spec
We analyze goals, constraints and KPIs together with the client's product team. We define scope, deliverables and acceptance criteria before estimating — no estimates on fuzzy scope.
Architecture
We design the data model, external integrations and contracts between modules. No code before the map is clear: you save weeks of downstream refactor.
Iterative development
Short cycles with weekly client demos, dedicated branch per feature, continuous code review. Every release is production-ready, not a throwaway prototype.
Review & test
Automated tests, QA checklist, security and accessibility audit before release. No surprises in production, no incidents in the first 48 hours.
Deploy & handover
Production deploy, operational documentation and training for your internal team for full post-project autonomy. You can continue with us or hand off with no hidden dependencies.