A PA system that has failed is indistinguishable from a PA system with nothing scheduled. Nothing beeps, no light turns red, and the first real test is the day it is needed.
The traditional answer is a walk round the building with a colleague and a two-way radio, once a term if anyone remembers. There is a better one.
Rather more than people expect, and it matters to be precise about which parts are genuinely verified and which are only inferred. Software can confirm that the audio device is still present, that scheduled files still exist, that the speech engine still produces audio, that each relay still switches, that remote rooms are still reporting in, and that what was scheduled actually played. What it cannot confirm is the last few metres:
What software alone cannot verify is whether sound left the cone. A blown driver, a disconnected speaker cable in a ceiling void, or an amplifier switched off at the wall is beyond it. For those, an amplifier with fault monitoring or a line-impedance check is the right equipment, and the honest thing to say is that the two are complementary rather than that one replaces the other.
A nightly self-test that nobody hears. Overnight, the device checks its outputs, its files, its speech engine, its relays and its rooms, and records the result. It is silent by design: a test that makes noise at 3am gets switched off within a week, and a supervision feature that gets switched off supervises nothing.
A weekly PA Health Report, printable, listing what was checked, what passed, what changed since last week and what needs attention. It exists to be filed and shown to somebody, because in most buildings the person who owns the answer is not the person who owns the system.
Alerts when something changes state, rather than a daily "everything is fine" message that trains people to delete it unread.
A history you can point at. When somebody asks whether the drill worked, the answer is a log entry with a time, not a recollection.
Once a term, at a sensible time: play a real announcement at real volume and walk the building. Automated supervision tells you the chain up to the amplifier is intact; ears tell you the room is covered and intelligible. Neither answers the other's question.
Does the self-test make noise?
No. It exercises the audio path without producing audible output. The relay self-test does click, which is why it is scheduled overnight and can be excluded on a site where that matters.
What if the device itself is off?
Then the report does not arrive and the room misses its heartbeat, which is itself the alert. An absent report is treated as a finding rather than as good news.
Can I get the results somewhere other than email?
Yes — the report is on the device's own dashboard, and the same state is available through the API, webhooks and MQTT, so it can feed a monitoring system you already run.
Does it detect a blown speaker?
No, and nothing that is only software can. It detects everything up to the amplifier input. Detecting the speaker itself needs monitored amplification, and a site that requires that should buy it.
How long is history kept?
Locally, until you clear it. The logs are plain text files on the device, so archiving them is a copy rather than an export.