Using GrowPilot
Anomaly detection (Cultivator)
Anomaly detection is the passive monitoring layer for Cultivator-tier collections. It looks at your scan history across plants and batches, identifies declining-health patterns (multiple plants trending unhealthy in the same week, a batch with worsening diagnoses, etc.), and surfaces them on the dashboard so you can intervene before the problem spreads.
What it watches
Cluster declines: when multiple plants in the same batch start scanning unhealthy within a short window.
Repeat diagnoses: the same disease appearing across multiple plants of the same species suggests a contagion path (shared water, shared tools, airborne spread).
Time-correlated unhealthy scans: even across different batches, if several plants flip from healthy to unhealthy within a few days, something environmental changed.
How alerts appear
Active anomalies show in a dedicated section on the dashboard with a brief description and a link to the affected batch or plants.
Push notifications fire when anomalies first trigger (if you've enabled Anomaly alerts in Settings). Email follows the same channel.
Anomalies auto-clear when the trend reverses, three consecutive healthy scans of the affected plants close out the alert.
How to act on an anomaly
Open the alert. The linked batch or plant cluster shows the recent scan history that triggered the alert.
Look for shared factors, same water source, same tray, same growing medium batch, recent environmental change (temperature swing, new fan, draft).
Treat preemptively. If the AI thinks 3 of your basil plants have early powdery mildew, the rest of the batch is at high risk. Treat all of them.
Frequently asked
- How quickly does anomaly detection react?
- Detection runs every few hours. The window for triggering is typically 5-7 days, long enough to filter noise, short enough to act before a systemic issue spreads further.
- Can I tune the sensitivity?
- Not currently from the UI. The thresholds are tuned conservatively to minimize false positives. If you're getting too many or too few alerts, contact support, we'll adjust based on your operation's profile.
- Does anomaly detection use my data to train AI models?
- Detection runs entirely on your data within your account, in real time. Your data is not used to train the underlying AI models. See the Privacy Policy for full data-handling details.