BeeSwarmly
  1. Home
  2. The Science
  3. Sensors & Tools
  4. Our Story
  5. FAQ
  6. Start Monitoring
From Sensors to Predictions

The Science.

How BeeSwarmly transforms raw hive data into actionable swarm warnings. Four stages, three sensor types, one clear output.

Sense, process, predict.

Each stage is automated. Each output is interpretable by any beekeeper.

1

Sensors collect continuously.

Temperature probes, a microphone, and a hive scale capture data every sixty seconds. The sensors run on long-life batteries and transmit wirelessly to a gateway device within 300 meters of the apiary. No cellular signal is required at the hive site.

2

Data flows to the cloud.

The gateway uploads sensor data over WiFi or cellular to our processing servers. Each reading is tagged with hive ID, timestamp, and environmental baselines. You see your data on the dashboard within seconds of collection.

3

The model calculates swarm probability.

Our prediction engine compares current readings against historical patterns from thousands of observed swarm events. It evaluates thermal gradients, acoustic frequency shifts, and weight anomalies to produce a swarm probability score updated every hour.

4

You receive an alert.

When swarm probability crosses your configured threshold, BeeSwarmly sends an alert via push notification, email, or SMS. The alert includes the specific hive, the predicted timeframe, and recommended interventions based on the detected trigger pattern.

Three data streams, one unified model.

Each sensor type catches a different phase of the pre-swarm sequence.

T

Thermal Array

Five temperature probes distributed across the hive body track brood nest temperature, cluster migration, and peripheral heat buildup. Queen cell construction raises local temperature by 0.5–1.2°C — a signal our model detects within hours of initiation.

A

Acoustic Sensor

A vibration-dampened microphone captures the colony's acoustic signature. Worker piping shifts from 200Hz to 350–500Hz in the 48 hours before departure. Our FFT-based classifier identifies these frequency changes with 89% accuracy in field conditions.

W

Precision Scale

A load cell rated to ±5g resolution tracks colony mass continuously. Swarming bees gorge on honey before leaving, producing a distinctive weight gain pattern in the 12–24 hours before departure followed by a sudden 1–3 kg drop at swarm event.

A false alarm wastes an afternoon. A missed swarm costs a colony. Our model is tuned to favor sensitivity over specificity.

Over two pilot seasons, our prediction engine correctly identified 91% of confirmed swarm events with an average lead time of 72 hours. The false positive rate sits at 8%, meaning roughly one in twelve alerts is a non-event. We consider that an acceptable trade-off: the cost of checking a hive that didn't need it is far lower than the cost of losing a queen and half a colony.

Every missed swarm teaches the model something new. Every false alarm does too. The system gets sharper with every season of data. — BeeSwarmly prediction principle

Want to see your hive data in real time?

Sensor kits ship within a week. Setup takes twenty minutes per hive. Your dashboard starts populating immediately. Join the pilot and see what your bees have been trying to tell you.

You will need A hive, WiFi within 300m, and twenty minutes
You will not need Technical expertise, cellular coverage at the hive, or a computer science degree