ANSI/ISA-18.2 operating guidance treats fewer than 6 alarms per operator per hour as a steady-state target, with fewer than 12 per hour considered manageable [S4].
These figures apply to normal plant operation, not the first minutes of a major upset, and they describe workload benchmarks rather than a universal pass-or-fail limit [S3][S10].
What alarm rate is acceptable under ISA-18.2?
The most useful engineering answer is fewer than 6 alarms per operator-hour for steady-state operation, with fewer than 12 per hour as the manageable ceiling [S4]. The frequently cited “one alarm per 10 minutes” target is equivalent to 6 alarms per hour, while a rate of one alarm every 5 minutes equals 12 alarms per hour [S3].
ISA-18.2-2016 did not retain the 150-alarms-per-day KPI used in the 2009 edition [S3]. That number would imply an average of 6.25 alarms per hour over a 24-hour period, but a plant shift is not interchangeable with a full day, and daily totals conceal short alarm peaks. The 1998 HSE CRR 166 benchmark specifically treated fewer than one alarm per operator per 10 minutes in steady state as “very likely to be acceptable” [S3].
Use fewer than 6 alarms per operator-hour in steady state as the initial design and acceptance target, then use 12 per hour as a review trigger. This is consistent with practical alarm-management guidance published on April 24, 2026 [S4].
Steady state and upset rates are not comparable
During the first 10 minutes after a major plant upset, fewer than 10 alarms per operator may be manageable, although complex responses make the load difficult [S3]. A total of 20 to 100 alarms per operator during that interval is difficult to cope with, while more than 100 is excessive and can cause the operator to abandon the system [S3].
These upset figures explain why a 12-alarm-per-hour benchmark must not be applied mechanically to an abnormal event. The relevant test is whether the operator can identify, prioritize, and act on the alarm within the required response time. ANSI/ISA-18.2-2016 defines the alarm’s primary function as notifying operators of abnormal process conditions or equipment malfunctions and supporting their response [S10].
Alarm priorities should therefore remain tied to consequence and response time, not simply to message order. A high-priority alarm should identify a condition requiring immediate action, while a low-priority condition may be better handled through operating procedures or a less urgent indication [S10].
Compare the main workload benchmarks
The main distinction is between acceptable steady-state load, manageable load, a short upset peak, and a daily legacy KPI. Each metric answers a different question, so mixing them produces misleading compliance conclusions. [S3]
| Benchmark | Equivalent rate | Decision use | Evidence |
| Very likely acceptable | Fewer than 1 alarm per 10 minutes, equivalent to fewer than 6 per hour | Steady-state design target | [S3] |
| Manageable | Fewer than 1 alarm per 5 minutes, equivalent to fewer than 12 per hour | Upper operating benchmark and rationalization trigger | [S3][S4] |
| First 10 minutes of a major upset | Fewer than 10 alarms per operator | Abnormal-condition workload guide | [S3] |
| Legacy daily KPI | About 150 alarms per day in 2009; about 144 per day in IEC 62682 | Historical context only, not the 2016 ISA-18.2 target | [S3] |
The comparison also shows why a site with 5 alarms per operator-hour can be acceptable while a site with 50 alarms during a ten-minute upset can be unmanageable. Rate must be evaluated together with duration, priority, alarm complexity, and required operator action.
How to measure the rate at an operating position
Measure annunciated alarms per operator during a defined operating interval and separate normal operation from upset periods. The working target is fewer than 6 alarms per operator-hour, and sustained performance above 12 per hour requires review [S4]. Use the same denominator for shift comparisons, because an 8-hour shift at 6 alarms per hour produces 48 alarms while an 8-hour shift at 12 per hour produces 96.
Track peak rate as well as average rate. A site can meet the hourly average while generating an unmanageable burst, and the HSE benchmark identified peak rate in any 10-minute period as a useful measure of variability [S3]. The 2026 SCADA implementation guidance also gives “one alarm per 10 minutes” as the steady-state target, no more than 10 alarms in a 10-minute peak period, chattering alarms below 5% of the total, and suppressed alarms below 1% [S5].
For operating positions, perimeter alarm and access control notifications should be evaluated separately from process alarms when they can generate large event bursts. The governing workload is the number of actionable indications presented to one person, not the raw event count across unrelated systems.
Who needs the 6-per-hour target and who does not
The target is for operators responsible for responding to process deviations or equipment malfunctions in a steady-state control-room environment [S10]. It is not a universal design limit for every indication, event log, interlock, or plant category. A single event can generate several indications if the system presents separate process, equipment, and system messages for the same abnormality.
The target is also less suitable for maintenance personnel, field operators, or a shared console where several people divide responsibility. Their actual workload must be assigned to the person expected to acknowledge and act on each alarm. For operators who also handle extensive surveillance camera review, security events should remain outside the process-alarm KPI unless the operator must respond to them.
Do not use 6 alarms per hour to justify retaining marginal alarms. If a point has no defined operator consequence, response time, or required action, rationalization should remove it, delay it, or convert it to another indication. The 2026 SCADA guidance describes rationalization as a line-by-line review of purpose, consequence, response, and priority [S4].
Why high alarm rates become a safety problem
Software-configurable automation can add alarms without the physical space constraints of a hardwired panel, but that capacity does not increase human response capacity [S2]. Alarm flooding can conceal the initiating alarm, delay corrective action, and contribute to environmental damage, production loss, injury, or death [S2].
A rationalization program commonly reduces active alarm counts by 50–80%, although the achieved reduction depends on the installed base and the condition of each alarm [S4]. Other failure signals include chattering, stale active alarms, and incorrect priority assignments [S5]. These are not solved by lowering the display threshold or increasing the alarm count target.
ISA-18.2 uses a 10-stage lifecycle covering philosophy, identification, rationalization, detailed design, implementation, operation, maintenance, monitoring, management of change, and audit [S4]. Rate is one output of that lifecycle, not a substitute for alarm-quality governance.
Standards basis and sourcing discipline
ANSI/ISA-18.2-2016 is titled “Management of Alarm Systems for the Process Industries,” and IEC 62682 provides the related international framework [S2][S10]. The ISA-18 series covers the design, operation, and maintenance of alarm systems used to promote safety and efficiency in process industries [S1].
For specification language, write “fewer than 6 alarms per operator-hour in steady state” as the target and “fewer than 12 per hour” as the manageable upper benchmark [S4]. Do not label 6 or 12 as a mandatory ISA-18.2 pass/fail limit. The 150-per-day figure belongs to the 2009 KPI history and was omitted from ISA-18.2-2016, while IEC 62682 cited about 144 alarms per day in 2014 [S3].
Before accepting a new operating target, document the measurement period, operating position, treatment of chattering and standing alarms, peak 10-minute load, upset exclusion rules, and escalation threshold. A trackable result is a sustained operator-hour rate below 6, a 10-minute peak rate below 72 alarms per hour, and documented rationalization of every alarm exceeding the agreed burden [S3][S4].