Fire Sprinkler Failure Rate in U.S. Building Fires, 2020 to 2025

Fire sprinkler failure rate in NFIRS building fires, 2020 to 2024

Across five complete NFIRS years, sprinklers failed to operate in 14.0 percent of the fires large enough to set them off and operated without controlling the fire in another 3.3 percent. The yearly rate moved within a narrow band, from 12.8 to 15.1 percent.

Large enough to activate: A sprinkler fire coded as operated and effective (AES_OPER 1), operated and not effective (2) or failed to operate (4). Fires coded too small to activate (3), other (0) or undetermined are not part of the rate. The NFIRS Complete Reference Guide defines effective as the system at least containing and controlling the fire until the fire department completes extinguishment.

From 2020 to 2024, fire departments reported 730,147 nonconfined building fires to NFIRS in our analysis universe (NFIRS Public Data Release). Sprinklers were coded present in 30,244 of them. In 10,491 of those sprinkler fires, the fire was too small to activate the system, which leaves 17,456 fires where the sprinklers had a chance to work.

Of those 17,456 fires, sprinklers operated and were effective in 14,440 (82.7 percent), operated but were not effective in 578 (3.3 percent) and failed to operate in 2,438 (14.0 percent). When sprinklers did operate, they were effective 96.2 percent of the time.

The 'as coded' rate counts every fire the reporting officer marked as a failure, including 1,041 fires where the failure reason was 'fire not in area protected by the system'. A sprinkler cannot be expected to control a fire it does not cover, so the table also shows a version without those fires: 88.0 percent operated and effective and 9.5 percent failed in 2020 to 2024. That second version lands within about one point of NFPA's published national estimates.

Which fires count: confined fires, building types and unknowns

AES fields are filled in consistently only for nonconfined building fires. Confined fires make up 50.3 percent of structure fires but only 29.3 percent of them carried a Fire Module, so they are counted separately and left out of every rate.

The brief covers structure fires, NFIRS incident types 111 to 123. Confined fires (types 113 to 118: cooking fires confined to a container, chimney fires, trash fires and similar) do not require the Fire Module, and the USFA analysis guidelines say AES data come entirely from that module and that confined fires should be removed from AES analysis. From 2020 to 2024 there were 792,455 confined fires; 232,484 had a Fire Module and 7,365 had an AES coded present.

The universe follows the USFA building-fire definition: incident types 111 and 120 to 123, structure type 1 (enclosed building) or 2 (fixed portable or mobile structure), mutual or automatic aid given removed so a fire is not counted twice, and every incident listed for exclusion in each year's errata addendum removed. Type 112 (fires in structures other than buildings) falls outside that definition.

Presence was coded undetermined in 9.9 percent of universe fires and no fires had it blank. Among fires with known status, any AES was present in 8.2 percent and a coded sprinkler system in 4.6 percent. A further 17,414 fires (32.4 percent of present-AES fires) had the system type left blank; the coding guide makes type required only when the fire was within the system's designed range, so these are read as fires the system was not positioned to fight and are not counted as sprinkler fires.

Wet pipe, dry pipe and special hazard systems

Dry pipe sprinklers failed to operate in 20.9 percent of large-enough fires, against 13.0 percent for wet pipe systems, a gap of 7.9 points.

Wet pipe systems were the system in 26,643 present-AES fires and operated and were effective in 84.0 percent of those large enough to activate. Dry pipe systems operated and were effective in 75.9 percent. Other sprinkler systems (deluge and pre-action) came in at 56.6 percent on a small base of 325 fires.

Special hazard systems (dry chemical, foam, halogen-type, carbon dioxide and other) had 1,711 large-enough fires between them. They operated and were effective in 55.5 percent and operated without effect in 13.9 percent, against 3.3 percent for sprinklers. The fire sprinkler inspection guide covers the water-based side, and systems served by a fire pump also depend on the checks described on the fire pump inspection page.

Why fire sprinklers fail: reasons ranked

Of 1,718 failed or ineffective sprinkler fires with a specific reason coded, 60.6 percent were fires outside the protected area. Setting those aside, conditions an inspection, testing and maintenance program is built to catch (shut-off systems, damaged parts, missed maintenance) account for 61.7 percent.

Sprinklers failed or were not effective in 3,016 fires large enough to activate them from 2020 to 2024. A specific reason was coded for 1,718; the rest were 'other', undetermined or blank (43.0 percent), a gap any reader of these figures should keep in mind.

Among failures to operate with a known reason other than coverage, the system had been shut off in 60.2 percent (513 known failures). Among sprinklers that operated without controlling the fire, the water did not reach the fire in 42.7 percent of known cases (164 fires). Manual intervention that defeated the system was 11.5 percent of known non-coverage reasons.

Building status matters. In buildings in normal use, sprinklers failed in 12.9 percent of large-enough fires. The rate was 30.7 percent in vacant and secured buildings, 61.3 percent in vacant and unsecured buildings, 33.5 percent under construction and 35.9 percent under major renovation.

Sprinkler performance by occupancy group

Apartments had the strongest record, with sprinklers operating effectively in 88.9 percent of large-enough fires; one- or two-family homes (75.1 percent), restaurants (71.7 percent) and schools (72.9 percent) were lowest among groups with 300 or more such fires.

Sprinklers were coded present in 12.6 percent of apartment building fires with known AES status, 49.7 percent of nursing home fires, 33.0 percent of manufacturing fires, 23.4 percent of warehouse fires and 0.4 percent of fires in one- or two-family homes.

Nursing homes had the highest sprinklered share of any group, yet sprinklers failed to operate in 19.5 percent of their 472 large-enough fires as coded (15.3 percent outside-area excluded). Inspection findings for those facilities are in our nursing home fire safety deficiencies report. Restaurants show the largest share of fires where sprinklers operated without effect (7.9 percent), a pattern NFIRS does not explain.

Hotels (80.7 percent), warehouses (81.7 percent), manufacturing (81.2 percent) and offices (77.4 percent) sit between those ends. Which buildings must have sprinklers at all depends on the code your state adopted; the requirement finder and the high-rise sprinkler retrofit laws report cover that side.

How many sprinkler heads opened

One head opened in 75.3 percent of operated sprinkler fires with a valid count, and five or fewer in 96.5 percent. Effectiveness drops as more heads open.

The NFIRS field for heads operating had a valid count (1 to 250) for 15,010 operated sprinkler fires. The median was 1 head when the system was effective and 2 when it was not; the 90th percentile was 3 heads vs 9.

When one head opened, the system was effective in 97.6 percent of fires; when 11 or more opened, in 80.5 percent. One head opened in 76.7 percent of wet pipe fires and 65.6 percent of dry pipe fires. The coding guide notes that the head count reflects how fast the fire grew and is not itself a measure of effectiveness.

Deaths, injuries and dollar loss with and without sprinklers

Building fires with sprinklers present had 2.41 civilian deaths per 1,000 fires, against 10.84 where no AES was present, 77.8 percent lower. This is an association inside NFIRS, not proof that sprinklers caused the whole difference.

In residential buildings the death rate was 3.46 per 1,000 fires with sprinklers present and 12.94 with none (73.3 percent lower). In nonresidential buildings it was 1.25 vs 2.43 (48.6 percent lower). Fire spread stayed within the object or room of origin in 84.7 percent of sprinklered fires and 46.2 percent of fires with no AES.

Two results run against the usual message and are reported as found. Civilian injuries per 1,000 fires were higher with sprinklers present (42.1 vs 33.4); 355 of the 1,272 injuries in sprinkler fires (27.9 percent) came from fires too small to activate the system. Mean dollar loss was also higher: $112,743 per sprinklered fire (112.7 million dollars per 1,000 fires) against $59,699 (59.7 million). The median loss points the other way, $3,500 vs $5,000, which suggests a few very large losses in big sprinklered industrial and storage buildings drive the mean. NFIRS dollar figures are fire officer estimates in nominal dollars.

How these counts compare with NFPA's national estimates

NFPA's U.S. Experience with Sprinklers (published 2026-10-01) covers the same 2020 to 2024 period with national estimates. Our raw counts land close to NFPA once fires outside the protected area are removed, and add the yearly series, the unknown shares, counted failure reasons and finer occupancy groups.

NFPA reports that sprinklers operated in 90 percent of fires large enough to activate them, were effective in 97 percent of those, and operated effectively in 87 percent, and that the death rate per fire was 90 percent lower with sprinklers than with no AES. NFPA estimates an annual average of 935 fires where the sprinkler system failed to operate and names a shut-off system as the most common reason.

The gaps have known sources. NFPA scales NFIRS to national totals with its own survey and excludes properties under construction; we report counts as filed, keep undetermined values visible and exclude only what USFA's errata require. Our sprinkler presence share (4.6 percent) is lower than NFPA's 11 percent partly because 17,414 present-AES fires with a blank type are not counted as sprinkler fires. Our death-rate gap (77.8 percent) is a raw ratio without NFPA's allocation of unknowns.

2025: the last NFIRS year is incomplete

USFA lists January 31, 2026 as the final date for NFIRS, which NERIS replaced. Building fires in the 2025 file fall to 52.7 percent of the 2024 level by December, so 2025 is shown as a partial check, not part of the trend.

The 2025 release notes report 21,366 departments and nearly 1,076,400 fire runs, against 22,582 departments and nearly 1,224,400 fire runs in 2024, a 12.1 percent drop in fire runs. In our universe, January 2025 building fires were 95.0 percent of January 2024; December 2025 was 52.7 percent of December 2024 (USFA on NERIS).

Rates from the partial year match the earlier years: sprinklers operated and were effective in 83.3 percent of large-enough 2025 fires and failed in 13.6 percent. Because 2025 is the last NFIRS release, future editions of this report will need NERIS data, which uses a different schema.

The numbers in one list

1. From 2020 to 2024, sprinklers failed to operate in 14.0 percent of 17,456 NFIRS building fires large enough to activate them (U.S., raw counts).^{[2][3][4][5][6]}
2. Sprinklers operated and were effective in 82.7 percent of those fires and operated in 86.0 percent (2020 to 2024, NFIRS).^{[2][3][4][5][6]}
3. When sprinklers operated, they controlled the fire 96.2 percent of the time (2020 to 2024, NFIRS).^{[2][3][4][5][6]}
4. Excluding fires coded as outside the protected area, the operated-and-effective rate was 88.0 percent and the failure rate 9.5 percent (2020 to 2024, NFIRS).^{[2][3][4][5][6]}
5. Sprinklers were coded present in 4.6 percent of NFIRS building fires with known AES status, 2020 to 2024; presence was undetermined in 9.9 percent.^{[2][3][4][5][6]}
6. Dry pipe systems failed in 20.9 percent of large-enough fires vs 13.0 percent for wet pipe systems (2020 to 2024, NFIRS).^{[2][3][4][5][6]}
7. A system shut off before the fire accounted for 60.2 percent of known failures to operate, outside-area fires excluded (2020 to 2024, NFIRS).^{[2][3][4][5][6]}
8. Fire not in the area protected was the most common coded reason, 60.6 percent of known reasons for failure or ineffectiveness (2020 to 2024, NFIRS).^{[2][3][4][5][6]}
9. Shut-off systems, damaged parts and lack of maintenance made up 61.7 percent of known non-coverage failure reasons (2020 to 2024, NFIRS; FLS grouping).^{[2][3][4][5][6]}
10. In vacant and unsecured buildings sprinklers failed in 61.3 percent of large-enough fires vs 12.9 percent in buildings in normal use (2020 to 2024, NFIRS).^{[2][3][4][5][6]}
11. In apartment building fires, sprinklers operated and were effective in 88.9 percent of large-enough fires; in nursing homes, 78.4 percent (2020 to 2024, NFIRS).^{[2][3][4][5][6]}
12. One sprinkler head opened in 75.3 percent of operated sprinkler fires with a valid count (2020 to 2024, NFIRS).^{[2][3][4][5][6]}
13. Civilian deaths were 2.41 per 1,000 building fires with sprinklers present vs 10.84 with no AES, 77.8 percent lower (2020 to 2024, NFIRS; association).^{[2][3][4][5][6]}
14. In the 2025 NFIRS file, December building fires were 52.7 percent of December 2024 as departments moved to NERIS.^[7][13]

Calculations

Download the calculations (CSV)

Summary

• Sprinklers failed to operate in 14.0 percent of 17,456 NFIRS building fires large enough to activate them, 2020 to 2024, and operated without effect in 3.3 percent.
• They operated and were effective in 82.7 percent as coded, or 88.0 percent when fires outside the protected area are excluded, close to NFPA's 87 percent national estimate.
• Wet pipe systems failed in 13.0 percent of large-enough fires, dry pipe systems in 20.9 percent.
• Inside the protected area, a shut-off system was the leading known cause of failure to operate (60.2 percent); shut-off, damaged and unmaintained systems together were 61.7 percent of known non-coverage reasons.
• Failure rates were far higher in vacant, unsecured buildings (61.3 percent) and buildings under renovation (35.9 percent) than in buildings in normal use (12.9 percent).
• Apartments showed the best record (88.9 percent operated and effective); one- or two-family homes, restaurants and schools the weakest among larger groups.
• Civilian deaths were 2.41 per 1,000 fires with sprinklers present vs 10.84 with no AES; injuries and mean dollar loss were higher, which we report as found.

Frequently asked questions

Related guides, services and research

• Fire sprinkler inspection requirements
• Fire pump testing requirements
• Requirement Finder: fire inspection checklist by building and system
• Fire Sprinkler License Requirements by State
• Firefighter Deaths per Year in the U.S., 2000 to 2025
• High Rise Sprinkler Retrofit Laws in the 50 Largest U.S. Cities
• Hotel Fire Safety Compliance

Methodology

Data: the FEMA/USFA NFIRS Fire and Hazardous Materials Public Data Releases for 2020 to 2025 (NFIRS Public Data Release), loaded unchanged into a read-only database. fireincident was joined to basicincident on INCIDENT_KEY and data year. Every code value used (AES_PRES, AES_TYPE, AES_OPER, AES_FAIL, PROP_USE, INC_TYPE, STRUC_TYPE, STRUC_STAT, FIRE_SPRD, AID) was checked against the codelookup table and the NFIRS 5.0 Complete Reference Guide, Chapter 5, Section M.

Universe: nonconfined building fires per the USFA analysis guidelines (INC_TYPE 111 and 120 to 123; STRUC_TYPE 1 or 2), aid given (AID 3 or 4) removed, and every incident each year's Data Quality Errata Addendum says to exclude from all analyses removed. Incidents the errata flag for dollar loss only were dropped from loss figures. Exposure fires were kept, as separate fires. Confined fires (113 to 118) were counted separately and excluded from all AES rates because their Fire Module is optional.

Sprinkler fires: AES_PRES 1 (present) and AES_TYPE 1, 2 or 3. Partial systems (AES_PRES 2) and present systems with a blank type were not counted as sprinkler fires; both counts are shown. 'Large enough to activate' means AES_OPER 1, 2 or 4. Rates exclude AES_OPER 3 (too small), 0 (other), U and blank from the denominator, and each excluded count is published. A second set of rates also removes fires with AES_FAIL 5 (fire not in area protected).

Failure reasons are AES_FAIL for sprinkler fires with AES_OPER 2 or 4. The ITM-addressable grouping (codes 1, 6, 7; possibly 2 and 3) is an FLS classification, not a USFA category. Occupancy groups come from PROP_USE codes (for example 419 one- or two-family, 429 multifamily, 311 nursing home, 891 warehouse, 700 series manufacturing). Head counts outside 1 to 250 were treated as invalid, following data quality notes that flag extreme values.

Outcomes: civilian deaths and injuries (OTH_DEATH, OTH_INJ, blank as zero) and dollar loss (PROP_LOSS + CONT_LOSS, nominal, fires with loss reported). Comparisons of sprinklers present vs no AES are associations inside NFIRS: sprinklered buildings differ in size, use, construction and occupants, and no adjustment was made for those differences.

Limits: NFIRS is voluntary and is not a census, so counts are 'fires reported to NFIRS', not national totals, and no state ranking is made. Effectiveness and failure reasons are the reporting officer's judgment. 2025 is incomplete because departments moved to NERIS during the year, so the trend stops at 2024 and 2025 is shown as a partial check. Conflicts with NFPA's published figures are shown side by side in table t8 rather than resolved; NFPA's figures are national estimates built with a different method.

Derived metrics are listed with formulas and input rows in derived_metrics.csv; all SQL is in the scripts folder and reruns against the public files. This report is a data summary and not engineering, legal or insurance advice; sprinkler requirements depend on the code edition adopted by your state and local jurisdiction. Last updated 2026-11-01.

This report presents published data with its limitations. It is not legal, engineering, financial or investment advice. Requirements depend on the code edition your state and local fire marshal have adopted; confirm with your authority having jurisdiction.

Sources, ranked by quality

Tier 1: primary and official sources (17)

• [1] FEMA, U.S. Fire Administration. FEMA/USFA NFIRS Annual Data (OpenFEMA data set page). Public Data Releases 1980 to 2025. Accessed 2026-11-01.
• [2] FEMA, U.S. Fire Administration, National Fire Data Center. NFIRS Fire and Hazardous Materials Public Data Release 2020 (fireincident, basicincident, codelookup, release notes, errata addendum). Incidents dated 2020. Accessed 2026-11-01.
• [3] FEMA, U.S. Fire Administration, National Fire Data Center. NFIRS Fire and Hazardous Materials Public Data Release 2021 (fireincident, basicincident, codelookup, release notes, errata addendum). Incidents dated 2021. Accessed 2026-11-01.
• [4] FEMA, U.S. Fire Administration, National Fire Data Center. NFIRS Fire and Hazardous Materials Public Data Release 2022 (fireincident, basicincident, codelookup, release notes, errata addendum). Incidents dated 2022. Accessed 2026-11-01.
• [5] FEMA, U.S. Fire Administration, National Fire Data Center. NFIRS Fire and Hazardous Materials Public Data Release 2023 (fireincident, basicincident, codelookup, release notes, errata addendum). Incidents dated 2023. Accessed 2026-11-01.
• [6] FEMA, U.S. Fire Administration, National Fire Data Center. NFIRS Fire and Hazardous Materials Public Data Release 2024 (fireincident, basicincident, codelookup, release notes, errata addendum). Incidents dated 2024. Accessed 2026-11-01.
• [7] FEMA, U.S. Fire Administration, National Fire Data Center. NFIRS Fire and Hazardous Materials Public Data Release 2025 (fireincident, basicincident, codelookup, release notes, errata addendum). Incidents dated 2025 (released 2026-05-01; incomplete, NERIS transition). Accessed 2026-11-01.
• [8] U.S. Fire Administration. NFIRS 5.0 Complete Reference Guide (2015), Chapter 5 Section M, Automatic Extinguishing System. Coding rules in force 2015 to 2025. Accessed 2026-11-01.
• [9] U.S. Fire Administration. NFIRS 5.0 Fire Data Analysis Guidelines and Issues. Analysis rules (building fires, confined fires, AES). Accessed 2026-11-01.
• [10] U.S. Fire Administration. National Fire Incident Reporting System (NFIRS) home page. Program status 2026. Accessed 2026-11-01.
• [11] U.S. Fire Administration. NFIRS Documentation (final date for NFIRS: Jan. 31, 2026). 2026. Accessed 2026-11-01.
• [12] U.S. Fire Administration. NFIRS Reporting Guidelines (2025 data deadline Jan. 31, 2026). 2024 and 2025 reporting years. Accessed 2026-11-01.
• [13] U.S. Fire Administration. National Emergency Response Information System (NERIS). 2026. Accessed 2026-11-01.
• [15] National Fire Protection Association (NFPA). U.S. Experience with Sprinklers (Tucker McGree, published 2026-10-01). 2020-2024 national estimates. Accessed 2026-11-01.
• [16] National Fire Protection Association (NFPA). NFPA 25, Standard for the Inspection, Testing, and Maintenance of Water-Based Fire Protection Systems (2026 edition). 2026 edition. Accessed 2026-11-01.
• [17] National Fire Protection Association (NFPA). NFPA 13, Standard for the Installation of Sprinkler Systems (2025 edition). 2025 edition. Accessed 2026-11-01.
• [18] National Fire Protection Association (NFPA). NFPA 13R, Standard for the Installation of Sprinkler Systems in Low-Rise Residential Occupancies. Current edition. Accessed 2026-11-01.

Tier 2: trade bodies, market research, company data (1)

• [14] Fire Safety Research Institute (UL Research Institutes). NERIS: The National Emergency Response Information System. Sole national system since January 2026. Accessed 2026-11-01.