Half the Testing, by Design

The Rule Change

A federal document cut fire alarm testing in half, and most of the people working under it have never opened it.

NFPA 72 is the national code for fire alarms. It calls for every device that starts an alarm, from smoke detectors to pull stations, tested once a year. Nobody picked that number at a conference table. Like most fire code, it came from old fires where the alarms stayed silent.

The military writes its own building standards, called the Unified Facilities Criteria (UFC). The fire protection volume carries the number UFC 3-601-02, dated 7 Oct 21 with Change 1 on 18 Apr 25. That volume tells users to follow its own schedules "in lieu of the tasks and frequencies in the National Fire Codes." Its Table 2-1 moves those same devices to a test every two years. Smoke detectors included.

Where the Numbers Came From

Nobody guessed at the two-year cycle either. It comes from a 1999 Air Force reliability study, Technical Report 01-10, written by the service's engineering support agency. The study put the chance of a fire at one event in fifty years. It put the chance a given test gets done right at 99 in 100. Run those numbers and a two-year cycle still comes out at 99 percent reliability.

On paper it reads clean. Half the labor for the same protection. I tested this gear in the field for twenty years, and the model itself holds up. The engineers who built it worked honestly with the assumptions they chose. The trouble lives in the assumptions.

What the Field Shows

I have stood on bases where six in ten fire alarm systems sat dead or partly dead. Missed maintenance piled onto missed reporting, season after season. The deficiency list, the standing list of known problems, became the only real inventory anyone had. The study promised 99 percent. The field returned numbers nowhere near it.

Look at what 99 percent asks of one tech on one ladder. He sprays a smoke detector with an approved test smoke, then reads how much smoke the device needed to trip. The code follows that reading against limits for five years. That is thinking work done with your hands. The record is a second task on the first, and it only counts if the next tech can use what got written down.

Cutting the test schedule in half raised the demands on the technician doing the testing, and no budget line ever paid for that.

The criteria expect a tech who handles the tools and the records in the same visit. The trade kept training hands-only workers while the study quietly redefined the job. Six in ten dead is what that gap grew into over twenty years.

What to Do Now

The people who wrote the document earn no blame from me. They built a model on stated assumptions, and the biggest assumption was the technician. It is printed in federal criteria for anyone to check.

The crews inside those buildings called it years ago. They kept saying the schedules fall apart under short staffing and turnover, and being right got them nothing. Field data was the one input the study never had. How long this equipment truly lasts is known only to the people who never stopped touching it. The deferred work is due now, and pricing the recovery falls to one labor pool.

So the fix is the boring kind. Train techs to the knowledge-worker standard the criteria already hold them to. Treat the test report as part of the building, because the next tech starts from whatever you leave behind. The whole plan rests on one person on a ladder. Train that person and pay him like it.

Next: the full review of Capitalism in America, the book behind this piece.