What a Fire Protection Record Has to Show a Year Later

Why this matters

Every fire protection system spends its whole life doing nothing visible. You cannot look at a sprinkler system and tell whether it was in service last March. The only thing that carries that answer forward is the record, which means the record is not paperwork about the system, it is part of the system. And it will be read by a stranger: a new facility manager, an insurance adjuster, an authority having jurisdiction, a lawyer, or you, twelve months from now with no memory of the visit. This card gives you the one question that separates a record from a checklist, then runs it against two real-shaped records.

The gate

Can a stranger, a year from now, with no access to anyone who was there, name the system's state on any given date, name who put it in that state, and name when it came back?

Three fields answer that, and a record that is missing any one of them cannot pass no matter how many boxes it has.

• State. In service, impaired, out of service, or degraded, expressed as a condition rather than as an activity. "Inspected" is an activity. "In service, all control valves verified open and sealed" is a state.
• Actor. A named person and their role. Not a company name, not initials nobody can resolve, not "tech."
• Clock. A start and an end. Any condition other than fully in service is a period rather than an event; the impairment card owns why that is true and the out-of-service SOP owns the procedure it triggers, and this card owns only the consequence, which is that both ends of the period belong on the record.

Where a measurement was taken, a fourth field applies: the value with its basis. A number without its instrument, its units, and the conditions it was taken under cannot be compared with next year's number, and comparison is the entire reason the number was taken.

Which instrument governs which record

None of these bind on their own. Each binds in the edition your authority having jurisdiction has adopted and amended, and the AHJ is a named role with authority to interpret and enforce that adopted edition, not a synonym for whoever inspects.

• Water-based systems - sprinkler, standpipe, private mains, fire pumps: NFPA 25 in the adopted edition, for inspection, testing and maintenance records and for the impairment program with its named impairment coordinator, preplanned and emergency impairment procedures, and tag.
• Fire alarm and signalling: NFPA 72 in the adopted edition, for the record of completion, the inspection and testing records, and the handling of a system out of service. Alarm system impairment is NFPA 72 territory, not NFPA 25; filing an alarm outage under the sprinkler impairment program is a common and consequential mix-up because the notification and compensating-measure requirements differ.
• Portable extinguishers: NFPA 10 in the adopted edition, for the monthly inspection record, annual maintenance, internal examination and hydrostatic test intervals.
• Opening protectives: NFPA 80 in the adopted edition, for the annual fire door inspection and its written record.

The federal floor is separate and much narrower, and it runs to your employees rather than to the building. 29 CFR 1910.157(e) requires annual maintenance checks on portable extinguishers with the date recorded and the record retained. 29 CFR 1910.164 covers fire detection systems and 1910.165 covers employee alarm systems, including test frequency for a non-supervised system. Never present a code figure as federal, and never present an OSHA figure as if it covered what a code covers.

Retention periods live in the adopted edition and commonly require you to keep the current record plus a stated number of prior ones, because several tests are only meaningful as a comparison. Confirm the number in the adopted edition rather than assuming.

One gate, two records

Same four-story building, same year, two documents in the same folder.

Record A: the contractor's annual sprinkler report. One page. Date, company name, a technician's first name, and a column of items each marked "satisfactory": control valves, gauges, main drain test, waterflow alarm, hangers, spare head cabinet. Signed.

Run the gate. State: absent. "Satisfactory" is an assessment of an activity on one day, not a condition. Actor: a first name that resolves to nobody after that person changes employers. Clock: one date, no periods. Value: the main drain test was the whole point of the visit and the record contains no static reading, no residual reading, and no description of the flow condition.

The failure is not that the tech did bad work. He may have done excellent work. The record simply cannot answer the question anyone will actually ask, which is whether the system was in service on the day something happened, and whether it was getting worse.

Record B: the same visit, recorded against the gate. Contractor report with numbers, plus the facility's own valve and impairment log.

• State: system in service at arrival, all four control valves verified open with seals intact and photographed; system placed on test 09:12 to 09:48 for the main drain and waterflow test, monitoring station notified before and confirmed off test after; in service at departure.
• Actor: technician by full name and license, facility engineer by name as the impairment coordinator under the adopted NFPA 25 edition.
• Clock: arrival, on test, off test, departure, all timed.
• Value with basis: main drain test at the system riser, 2 in main drain fully open, gauge identified by serial number with its calibration date, static 74 psi, residual 52 psi. Prior year, same gauge and same drain: static 76 psi, residual 58 psi.

That record passes the gate, and because it passes, it produces a finding the checklist could not.

What the numbers say, and what would make them say nothing

This year's drop at flow is 74 minus 52, which is 22 psi. Last year's was 76 minus 58, which is 18 psi. The drop grew by 4 psi while static fell by only 2 psi.

Error basis and character, applied before the comparison. Take an illustrative field gauge whose stated accuracy is 3 percent of a 300 psi span. That is 9 psi anywhere on the dial, its basis is percent of span rather than percent of reading, and its character is a worst-case bound bundling zero offset, span error, linearity and hysteresis. Only the zero-offset component cancels when you difference two readings on the same gauge; linearity and hysteresis between 74 psi and 52 psi are two different points on the dial and do not cancel, so the surviving term is repeatability plus the manufacturer's linearity spec, read off the datasheet rather than assumed to be zero. A fixed offset from ONE gauge cancels in a difference taken with that gauge, so the within-year drop of 22 psi does not carry the 9 psi at all. What survives is the repeatability, a small independent random term; at an illustrative 1 percent of reading that is 0.74 psi at static and 0.52 psi at residual, and independent random terms combine in quadrature, giving about 0.9 psi on this year's drop and about 0.95 psi on last year's. The two years combined in quadrature give about 1.3 psi on the 4 psi change. Four is comfortably outside 1.3, so the change is real.

What would erase the finding. If the two years had been read on different gauges, the fixed offsets no longer cancel between years. Two 9 psi bounds add linearly and are written with one inequality sign, so the between-year comparison would carry an offset difference of not more than 18 psi. A 4 psi change inside a not-more-than-18 psi bound means nothing at all, and no amount of trending recovers it. That is why the gauge identity is a required field and not a detail: it is the difference between a number that supports a conclusion and a number that cannot.

What the finding points at, with a direction check. A partially closed control valve, an obstruction, or a closed sectional valve raises friction between the supply and the riser, which shows up only under flow. It therefore raises the drop while leaving static nearly unchanged, which is exactly the pattern here. At the other end of the range, a fall in the municipal supply pressure lowers both static and residual by similar amounts and leaves the drop unchanged. Those two causes are distinguishable on this record and indistinguishable on Record A.

How the test was performed, because it is not a benign act. Opening a main drain discharges a large volume of water. The discharge was routed and observed so it did not erode the grade, flood the loading dock, or sheet the exit discharge path with ice, which would have taken an egress path out of service while proving a sprinkler system was in one. The valve was opened and closed slowly to avoid water hammer on the riser. The system was placed on test before the flow so the waterflow alarm did not dispatch the fire department, and taken off test with a confirming callback. No control valve was closed at any point to "check" it; closing a control valve to any degree is an impairment under the adopted NFPA 25 edition and starts the impairment procedure with its tag, its compensating measures, and its named coordinator. Nothing was unscrewed from a pressurized port; a gauge plug at riser pressure ejects.

Sibling-rule check. The discharge path was kept off the exit discharge, which the egress card requires and which the ice case would otherwise have violated. The on-test and off-test times carry an actor and a confirming callback, which is the same discipline the hot work SOP applies to a bypassed detection zone. The alarm system's out-of-service handling is attributed to the adopted NFPA 72 edition and the sprinkler impairment to the adopted NFPA 25 edition, not merged. Every level here carries its instrument, its units, and its flow condition.

The record defects that survive an audit

Three shapes pass a structural check while failing the gate, and they are worth knowing by sight.

The rolled-forward record. Last year's form with the dates changed. The tell is identical readings across years on a system that is a live hydraulic object; a main drain residual that is the same number three years running is either a copy or a gauge that stopped moving, and both are findings.

The open-ended impairment. A start time, a tag, compensating measures, and no end. Whoever restored the system did not write it down, so the record says the system is still impaired. A stranger reading it a year later has to conclude the building has been unprotected since, and an adjuster will.

The unresolved actor. Initials, a first name, or a company. When the person is needed, they cannot be found, and the record loses the one thing that made it evidence rather than an assertion.

References

• NFPA 25, Standard for the Inspection, Testing, and Maintenance of Water-Based Fire Protection Systems, in the edition adopted and amended by your authority having jurisdiction, for ITM records, comparison with previous results, and the impairment program
• NFPA 72, National Fire Alarm and Signaling Code, in the adopted edition, for records of completion, test records, and system out-of-service handling
• NFPA 10 for portable extinguishers and NFPA 80 for fire door inspection records, in the adopted editions
• 29 CFR 1910.157(e) for extinguisher maintenance records, 1910.164 for fire detection systems, and 1910.165 for employee alarm systems
• See related: What Emergency and Exit Lighting Have to Do; Running a Hot Work Permit That Would Hold Up