A fire alarm tells you something is wrong. It doesn’t tell you what to do.
For decades that was sufficient, because there was only one answer. Alarm means leave the building. Everybody knew it, everybody practiced it twice a year, and the response was the same regardless of what had actually triggered it. The system worked because the question only ever had one answer.
That assumption no longer holds. Emergency planning now asks organisations to distinguish between evacuation, invacuation, lockdown and shelter in place. Those responses are not variations on a theme. They are opposites. Evacuating during a lockdown moves people towards the threat. Locking down during a fire traps them inside it. A single tone cannot carry that distinction, and asking people to infer the right response from a sound puts the decision in the worst possible place: with whoever happens to be standing nearest, holding the least information.
An alert says something is happening. An instruction says what to do about it.
That is the whole distinction, and almost every problem with emergency communication follows from collapsing the two together.
An alert is a signal. It draws attention, raises the pulse, and interrupts whatever people were doing. It is genuinely valuable, and for a straightforward fire, it may be all you need, because the alert and the instruction happen to be the same thing.
An instruction is different. It tells a specific group of people to take a specific action, right now. It doesn’t require interpretation, it doesn’t depend on the recipient guessing the nature of the incident, and it doesn’t assume everyone has read the same page of the same policy document at the same time.
Most organisations have invested heavily in the first and barely at all in the second.
Why ambiguity costs time
There is a common assumption that people react to alarms immediately. They don’t.
People respond to an ambiguous signal by seeking confirmation. They look at colleagues. They check whether anyone else has stood up. They glance out of the window, ask the person next to them whether this is a drill, and wait for someone with visible authority to say something definite. Fire engineers call the resulting delay pre-movement time, and the milling behaviour that fills it is one of the most consistently observed findings in evacuation research.
None of that is irrational. It is entirely sensible behaviour when the signal itself is vague. The problem is that it happens at the start of an incident, when they time is most valuable, and it scales badly: the larger the group, the longer people spend looking at each other.
A clear instruction collapses that gap, because there is nothing left to interpret. “Lockdown now. Lock the door, move away from windows, stay quiet.” Removes the confirmation-seeking step entirely. There is no ambiguity to resolve, so nobody stands still resolving it.
What a good instruction actually contains
The action, named. Not “there is an incident on site” but what to do about it. An alert that conveys only urgency hands the interpretation problem straight back to the recipient, which is exactly what you were trying to avoid.
The people it applies to. Sometimes everyone needs the same message. Often, they don’t. The people in the affected building may need to lock down while everyone else needs to stay away from it, and a single undifferentiated broadcast can’t do both.
Coverage. A message delivered only to the places people sit is an alert to everyone else. Car parks, outbuildings, open ground, stairwells and the far end of a split site are where instructions most often fail to land, and they’re rarely where anyone tests them.
Pre-written wording. Nobody composes calm, unambiguous sentences under pressure. The people best placed to write good emergency wording are the people who wrote your procedures, working slowly, months in advance, with time to argue about the phrasing. Doing it live, one-handed, with adrenaline up, produces exactly the vague message this post is about.
The message nobody plans for
Ask an organisation how a lockdown ends and you will usually get a pause.
The stand-down tends to happen by someone walking the building and telling people in person. On a large site that takes a long time, and until it’s finished you have people sitting behind locked doors with no information, which is its own kind of harm.
An alarm has no vocabulary for ‘it’s over, resume normal activity’. It can only be on or off, and switching it off communicates nothing except that the noise has stopped. The all-clear is a useful diagnostic when you’re assessing your own setup, because it’s the one message an alarm-based system fundamentally cannot send.
Where this meets Martyn’s Law
The Terrorism (Protection of Premises) Act 2025 applies at the standard tier from a capacity of 200, and the enhanced tier from 800, with commencement expected in spring 2027.
The standard tier duties are procedural. They concern what people are told to do. Evacuation, invacuation, lockdown and communication. Rather than what equipment detects. That maps almost exactly onto the distinction this post is about. The legislation is interested in whether you can get the right instruction to the right people, not in whether you can make a noise.
No supplier can sell you compliance, and it’s worth being wary of anyone who suggest otherwise. What a system can do is narrow the gap between the procedure you’ve written down and your ability to actually deliver it when it matters.
The practical implication
If your current system can only make one sound, then in practice it can convey on instruction. Every other scenario in your emergency plan is quietly relying on somebody with a mobile phone and a mental list of who to ring.
That isn’t an argument for replacing your fire alarm. Fire alarms are very good at the job they were designed to do, and that job hasn’t changed. It’s an argument for recognising that alerting and instructing are two separate tasks, and that most organisations have only ever built for one of them.
