According to the National Fire Protection Association's latest research, U.S. fire departments responded to an estimated annual average of 1,544 structure fires at warehouse properties between 2020 and 2024 — causing two civilian deaths, 17 injuries, and roughly $314 million in direct property damage each year. As warehouses get bigger and fuel loads grow, the bill for a single fire keeps getting harder to bear.
And yet, ask a simple question — "how often should a warehouse run an evacuation drill?" — and no federal regulation gives you a single, clean answer. This piece follows the data to find where the answer actually lives.
The Regulations Don't Give You the Number
Let's start with what the rules actually say. OSHA's 29 CFR 1910.38 requires employers to maintain a written Emergency Action Plan (EAP) — fire reporting procedures, evacuation routes and assignments, post-evacuation accounting, rescue and medical duties. But the standard does not specify how often drills must be held. OSHA's own guidance phrases it as holding practice drills "as often as necessary to keep employees prepared." That is a recommendation, not a mandate.
The International Fire Code, Section 405, Table 405.2, sets drill frequencies by occupancy: factory/industrial occupancies (Group F) must drill at least annually, with employee participation. Notably, at the model-code level, storage occupancies (Group S) — which is what most warehouses are — sit outside the mandatory drill list. Whether that exemption survives in practice depends entirely on the local authority having jurisdiction; in real operations, most warehouse facilities are pulled into local drill requirements anyway. Don't read the exemption as a free pass.
NFPA 101, the Life Safety Code, similarly sets no explicit drill frequency for industrial or storage occupancies.
So the honest first answer to "how many times a year" is: there is no single federal number. The working industry consensus — quarterly for higher fire-risk sites, semiannually for ordinary-risk sites — is custom, not code.
This regulatory gap explains why drills so easily become theater: when the law sets no clear frequency bar, "we did one once" and "we drill every year" look nearly identical on a compliance checklist, and box-checking gets institutional cover.
What the Data Says About the Most Dangerous Moments
Two data points deserve attention.
First, the NFPA warehouse fire profile: fires between midnight and 6 a.m. account for just 18 percent of incidents but 40 percent of direct property damage. Almost every drill happens in daytime, with full staffing — ideal conditions. Fires, it turns out, prefer the worst conditions: thin night shifts, longer decision chains, slower outside response. Drills always test the A-side; real fires grade the B-side.
Second, a concrete case. In the early hours of December 18, 2021, a QVC distribution center in Rocky Mount, North Carolina, caught fire. Roughly 75 percent of the 1.5-million-square-foot facility was destroyed — the largest structural fire in state history. More than 300 employees were on site. Officials first said everyone had been accounted for, then discovered one person missing: 21-year-old Kevon Ricks, whose body was found near the outbound dock — in a section the fire had never reached. Investigators believe smoke inhalation was the cause. After a yearlong investigation, the ATF listed the fire's cause as undetermined.
The detail warehouse operators should sit with is not how big the fire was, but what happened in the minutes after evacuation: an inaccurate headcount can lead responders to conclude nobody is left inside. And the headcount is precisely the part of a drill most likely to be faked — everyone stands at the muster point chatting, the supervisor shouts "all here?", someone shouts back "all here," drill over.
Why Drills Become Theater
Across many drill observations, the causes of formalism cluster into four.
First, misaligned goals. The drill's KPI is set as "evacuation time," so everyone optimizes for speed. But speed is only a quarter of a drill's value. The other three quarters: does the backup route actually work, is the muster-point headcount accurate, does the reporting chain function. A drill that only measures speed cannot surface the problems that actually kill.
Second, permanently gentle scenarios. Drills happen in daylight, with the main aisle clear and lighting intact. Nobody blocks the primary route. Nobody schedules one at 3 a.m. during a shift change. Yet NFPA data shows the overnight hours carry the heaviest losses.
Third, incomplete coverage. Night-shift workers, temps, outsourced forklift operators, delivery drivers, visitors — the people least familiar with evacuation routes — are routinely absent from drill rosters that list only day-shift regulars.
Fourth, no debrief loop. The drill ends and the work ends. No timing records, no issue list, no owners, no deadlines. The same problems can repeat across two drills in the same year and nobody notices.
Four Dimensions of a Drill That Works
Once the causes are clear, the design framework converges on a simple scorecard. If a drill is just "walking through it," grade it against these four dimensions:
| Dimension | Design requirement | How to measure |
|---|---|---|
| Scenario realism | Block one primary route each time; run at least one night or overnight drill every two years; simulate power-out, no-lighting conditions | Backup-route utilization rate; time gap between night and day drills |
| People coverage | Night shifts, temps, contractors, visiting drivers all on the roster | Muster-point accounting coverage (expected vs. actual) |
| Timing and headcount | Time the full sequence from alarm to completed headcount | Total headcount completion time; headcount error (difference between first count and verification) |
| Debrief closure | Written debrief within 48 hours; every corrective action has an owner and a deadline | On-time closure rate of corrective actions |
Here's the counterintuitive call: the most valuable output of a drill is not "how many minutes the evacuation took" but the headcount error. The QVC case shows what a headcount error can cost. Make "zero headcount error" the hard metric — not "evacuate faster."
And a trade-off worth pricing: a 200-person drill costs roughly an hour of operations — a direct, visible cost. A botched evacuation costs a probability-weighted catastrophe. The real comparison isn't "drill cost vs. zero," it's "drill cost vs. higher insurance premiums, lost points in customer audits, and a low-probability but ruinous accident." Put the drill inside that equation and it stops looking like pure expense.
The Answer Is Moving From Regulation to Contract
One last trend read.
Federal rules are likely to stay vague on drill frequency for the foreseeable future. But two harder constraints have emerged in recent years: insurers, who increasingly tie premiums and coverage terms to drill records; and brand customers auditing their 3PLs, whose supplier scorecards now grade drill frequency, records, and debriefs. As warehouses grow larger and automation spreads — high-density ASRS storage makes fires markedly harder to fight — both constraints will only tighten.
In other words, the answer to "how many times a year" will be written less in regulations and more in insurance contracts and customer agreements. Warehouses still waiting for the law to give them a number may find the question arrives first at renewal time — or at the next customer audit.
When was your warehouse's last drill? What did it measure? What did it change? If you can't answer two of those three, that drill was probably theater — and the difference between a theatrical drill and no drill at all may be smaller than you think.
By Dr. Chen, Supply Chain Researcher
