NFPA publishes a single headline figure from its report on the U.S. experience with sprinklers: sprinkler systems control fire successfully in 89% of cases. The industry repeats it. Few ask what it actually measures.

The methodology

The denominator is fires in which the system activated. Cases where the system never started are excluded by definition — NFPA acknowledges this; the headline does not. Within the activation failures, the report attributes 61% to a closed control valve — human error during maintenance — and 39% to technical causes: pump, mains, pressure loss. But that is the structure of failure, not its scale. The total number of systems that failed to activate is never published, because those cases sit outside the 89%. How many systems failed is unknown; how many succeeded is equally uncertain, since minor fires controlled quickly are under-reported in NFIRS. The figure is calculated from an unknown numerator over an unknown denominator.

The standard does not verify readiness

NFPA 25 does not require live fire testing. The sprinkler head is a sacrificial component — it destroys itself on activation — and therefore cannot be tested in place. The maintenance regime checks pipe pressure and pump start-up; whether a specific head will still activate at the correct temperature ten years after installation is verified by no standard. The system is certified ready on indirect indicators. Its actual readiness is an assumption, not a measured fact.

The environment changed; the architecture did not

Centralised architecture was designed for stable peacetime infrastructure — municipal water, grid power, trained personnel, regular servicing. A single strike on the pump room, the reservoir, or the main riser removes fire protection for the entire building. This is not a theoretical scenario: it has been documented across Ukraine, Russia, and increasingly the Gulf over the past two years.

One link breaks — the whole system fails.

NFPA 13 has not changed this architecture. The threat environment has changed entirely.

A different architecture — not a competing product

RadEX Water Stream is built the other way round. An autonomous water-mist module has no pump, no mains, no riser, no valves, no electrical supply, and no central control panel. Inside the cylinder there is a suppression agent, a pressure generator, and a thermal trigger. When the local temperature exceeds the activation threshold, the trigger fires the generator, and water is forced through the nozzle and converted to fine mist — with no external resources and no external points of failure.

Each module is independent: the failure of one does not affect any other, because there is no mechanical or hydraulic connection between them. There is no central point, so the system cannot be shut down centrally. The cylinder is not pressurised in standby — pressure is generated only at activation. A direct strike on the room, in most realistic scenarios, tears the module from its mount; it activates in any orientation, including on its side under debris.

Readiness that can be tested

One module, chosen at random from an installation, is activated under real conditions. That is the test — not a bypass-valve flow check, not a visual inspection. The result is observable, measurable, and repeatable. This is what verification of readiness actually looks like.

RadEX Water Stream — autonomous water-mist module under test.

The position

We are not proposing to remove existing deluge systems. The statement is narrower and factual: the operating environment has changed, centralised architecture no longer matches it, and the standard used to measure effectiveness does not measure what matters. An 89% control rate — in systems that activated, drawn from voluntary incident reports in peacetime conditions — measures how well a system performs when everything upstream of the head is already working.

Where a single strike removes the entire upstream chain, that number answers nothing.