# Reducing False Downtime Alerts for Smart Factories

> TechSupport Team supported the Kafka and TimescaleDB alert pipeline on FabriSense’s factory monitoring platform. Over five months, client Datadog records show false downtime alerts falling from 240 to 77 per week. Notification time fell from nine minutes to 90 seconds.

- Publisher: TechSupport Team
- Canonical page: https://techsupport-team.com/case-studies/smart-factory-iot-alert-calibration/
- Updated: 2026-11-16

Client: FabriSense · Telecom & Manufacturing · German smart-factory IoT provider (180 staff)

## Client and engagement facts

- **Client:** FabriSense
- **Industry:** Telecom & Manufacturing
- **Business context:** German smart-factory IoT provider (180 staff)
- **Delivery model:** IoT support
- **Application:** Smart-factory IoT with downtime alerting
- **Supplied geography:** German
- **Company size in source:** 180 staff

## Application stack and tools

### Application stack

- Python
- Kafka
- TimescaleDB
- Kubernetes
- Azure IoT

### Tools used

- Datadog
- PagerDuty
- Grafana

Related services: [Application support](https://techsupport-team.com/services/application-support/)

## Why Support Was Needed

Shared thresholds across device types and Kafka consumer lag triggered false downtime alerts that interrupted production. The engineers tuned thresholds by device model and addressed lag in the pipeline.

## Technical Work

- Added sensor threshold checks per device model.
- Fixed the Kafka consumer lag causing 40% of false alerts.
- Built per-device alert calibration.

## Representative Incident

A plant network outage generated a cluster of false alerts in month three. TechSupport Team recorded the impact in the alert register. The client retained responsibility for plant networking.

## Results and how they were measured

Table: Reported results and their measurement scope

| Metric | Baseline | Result | Units | Measurement window | Source and scope |
| --- | --- | --- | --- | --- | --- |
| False alerts | 240/wk | 77/wk | /wk | 5 months | Client Datadog; Scope: The five-month false-alert comparison excludes one plant network outage. |
| Notify time | 9 min | 90 s | Baseline: min; Result: s | 5 months | Client PagerDuty |
| Dashboard uptime | 99.1% | 99.9% | % | 5 months | Client Grafana |

## Evidence and Limitations

The five-month false-alert comparison excludes one plant network outage. Notification time and dashboard uptime use PagerDuty and Grafana.

## Plant and Platform Ownership

FabriSense managed on-site networks. TechSupport Team handled alert calibration and the data pipeline.

## Scope a Similar Engagement

Share which devices generate false alerts and where plant networking hands off to the application team. [Request a support plan](https://techsupport-team.com/contact/) or explore [application support](https://techsupport-team.com/services/application-support/).

## Related pages

- [Application support](https://techsupport-team.com/services/application-support/): The related service for this engagement.

## Contact

Request engineers from TechSupport Team or ask a question: https://techsupport-team.com/contact/
