FlockIQ — IoT Monitoring for Poultry Operations

Industry: AgTech
Service: Full product build
Timeline: 10 weeks
Result: 19-sensor IoT system. Alert response time reduced from 4 hours to 90 seconds.

---

The problem

A mid-sized poultry operation was monitoring flock health manually. Staff would walk the houses every few hours, checking temperature, humidity, and CO₂ levels. If something went wrong between checks — a heater failure, ventilation blockage, or spike in ammonia — it could mean hours of exposure before anyone noticed.

The client had looked at off-the-shelf IoT monitoring systems. Every option required expensive proprietary hardware, long contracts, and didn't integrate with their existing SCADA data. They needed something built for their specific operation.

What we built

A full IoT monitoring platform: hardware integration, real-time data pipeline, alerting system, and dashboard.

Hardware integration: 19 sensors across 4 houses — temperature, humidity, CO₂, and NH₃. Sensors pushed data to a central hub every 30 seconds via MQTT.

Data pipeline: Node.js ingestion service, TimescaleDB for time-series storage (Postgres extension — no new infrastructure). Data retention policies to manage storage costs.

Alerting system: Threshold-based alerts with configurable sensitivity by sensor and time of day. SMS and email via Twilio and Resend. Alert escalation if primary contact doesn't acknowledge within 10 minutes.

Dashboard: Next.js frontend showing real-time sensor readings, historical charts, house-by-house comparison, and alert history. Mobile-responsive for on-site use.

Admin panel: Add/remove sensors, adjust thresholds, manage contacts, view audit logs.

The technical decisions

TimescaleDB over InfluxDB: The client already had PostgreSQL experience and a Postgres DBA. TimescaleDB gave them time-series performance without new tooling or a new database to manage. Compression reduced storage costs by ~80% vs. raw Postgres.

MQTT over HTTP: Sensors were in areas with intermittent connectivity. MQTT's lightweight protocol and retained message support handled disconnections gracefully. HTTP would have required retry logic at the sensor level.

Alert deduplication: A naive implementation would fire 20 alerts in 10 minutes when a sensor started drifting. We implemented alert deduplication with a configurable cool-down period and "all-clear" notifications when conditions returned to normal.

Results

• Alert response time: 4 hours (manual walk) → 90 seconds (SMS alert)
• Sensor coverage: 8 sensors (manual spot checks) → 19 continuous sensors
• False positive rate: Less than 2% of alerts after threshold calibration
• Uptime: 99.8% over the first 6 months

The client's operations manager said the system paid for itself in the first month when it caught a heater failure at 2am that would have been a full flock loss by morning.

Handoff

The full codebase, infrastructure configuration (Railway + Vercel), sensor setup documentation, and a recorded walkthrough were delivered at week 10. The client's internal team manages the system independently, with a Silver Owl retainer for annual maintenance and any new house additions.