An Alarm Nobody Trusts Is Worse Than No Alarm
Every failed system we investigate remotely shares one symptom: the alarms had been going off for a while and had stopped meaning anything. Too many alerts, too many false ones, nobody had reviewed the thresholds since commissioning, and the one that mattered came in at 02:00 alongside eleven others.
A useful hydroponic alarm system is not the one with the most alerts. It is the one where each alert means someone specific should do something specific, and where that rarely happens for no good reason. This guide covers what to alarm, how to set thresholds, who should be contacted and in what order, and how to keep the system credible.
What Deserves an Alarm
| Condition | Why it alarms | Typical priority |
|---|---|---|
| Pump failure or loss of flow | Crop can lose its water supply within hours | Highest |
| Power failure | Stops pumps, dosing and controls simultaneously | Highest |
| Tank level low or overflow | Runs pumps dry or floods the plant room | High |
| EC or pH out of range | Nutrition drift that compounds quickly | High, with delay |
| Temperature out of range | Crop damage or growth rate change | High, day and night |
| Dosing not running when expected | Silent failure until the crop shows it | Medium–high |
| Humidity or VPD extremes | Disease and physiological disorder pressure | Medium |
| Sensor fault or out-of-range signal | A failed sensor disabling your protection | Medium, treated seriously |
| Communication loss | The system may be blind to everything else | Medium–high |
A helpful test: if an alert arrives and no action is expected, it is a notification, not an alarm. Put it in the daily report instead.
Setting Thresholds
- Start from crop tolerance, not from round numbers. The threshold should sit before damage, not at the damage point
- Use delay timers. Transient spikes at pump start-up, dosing events and door openings otherwise generate false alerts every day
- Differentiate day and night limits where the crop’s safe range genuinely differs
- Add deadbands so alerts do not chatter around a setpoint
- Document the rationale for each threshold so that whoever tunes it next year understands why it is there
- Review after incidents. The correct time to adjust a threshold is after you have seen how it behaved
Escalation: Who Gets Called and When
- Primary contact — the person on shift or on call, notified immediately for high-priority alarms
- Second contact after a defined no-response interval, because phones go flat and people sleep deeply
- Third contact or owner after a longer interval for anything that can destroy a crop
- Supplier or service contact for conditions that need technical support rather than site action
Escalation that relies on one person responding fails eventually. Build the ladder, test it, and keep contact details current — a stale contact list is the most common reason a working alarm system achieves nothing.
Notification Channels
| Channel | Strength | Limitation |
|---|---|---|
| Audible alarm on site | Immediate for anyone present | Useless when nobody is there |
| SMS / text | Reaches people without data coverage | No context; easily ignored when frequent |
| App notification | Rich information, acknowledgment tracking | Requires data and app discipline |
| Phone call | Hard to miss | Only justified for the highest priorities |
| Email digest | Good for trends and low-priority items | Too slow for anything urgent |
Use more than one channel for the critical tier. Redundancy in notification is cheaper than redundancy in equipment.
Avoiding Alarm Fatigue
- Fewer, better alarms. Every nuisance alert reduces the response probability of every other alert
- Review the monthly log and either fix the cause or remove the condition from the alarm list
- Separate alarm from trend. Trends belong in reports reviewed calmly; alarms belong to conditions requiring action
- Re-tune after changes. New equipment, new hardware or a new crop changes what normal looks like
- Test deliberately and regularly so that everyone knows what a real alarm looks like
- Acknowledge and log responses. Tracked acknowledgment is what separates a monitoring system from a noisy one
Redundancy That Actually Helps
- Duplicate critical sensors rather than trusting one probe to protect everything
- Independent flow detection rather than inferring flow from pump run status
- Uninterruptible supply for controllers so alarms survive the power failure they are reporting
- Local audible backup independent of the networked system
- Fail-safe design — decide what a system does when it loses signal, and make sure that state is the safe one
Commissioning and Routine Testing
- Trigger every alarm individually and record who was notified and when
- Test the escalation ladder by letting the first interval expire deliberately
- Confirm the system reports a simulated power failure
- Verify sensor readings against calibrated references before trusting thresholds
- Repeat quarterly, and after any change to equipment or staffing
- Keep the results — they are also evidence of due diligence after an incident
FAQ
How many alarms is too many?
When staff can no longer say what each one means or what to do about it. If a condition has no defined action, it belongs in a report rather than an alarm.
Should low-priority events alarm at night?
Rarely. Tier them: critical conditions interrupt sleep, everything else waits for the morning report. Otherwise critical alerts lose their power.
What causes most false alarms?
Thresholds with no delay timer, duty cycles that momentarily exceed limits, dirty or drifting sensors, and poorly located probes measuring transient conditions.
Do I need redundancy with a cloud system?
Yes. Cloud monitoring depends on connectivity that can fail. Keep a local audible or visual indication that works regardless.
How often should we review thresholds?
At least annually, and immediately after any incident or season change. Most commercial systems are set at commissioning and never touched again.
Can alarms help with audits and insurance?
They can. Logged alarm history with acknowledged responses demonstrates active management, which matters for both certification and claims.
Make Every Alert Worth Responding To
Alarm logic belongs in the design stage, where it costs nothing to add the right sensor or escalation path. Send us your system layout and staffing pattern and we will specify monitoring points and alarm tiers. Reach us through the quote form.
Related reading: alarms connect to everything else running the farm — see IoT and remote monitoring, risk management and power failure, sensor calibration, electrical safety and backup and multi-site reporting and remote control.