Compliant on Paper: The gap between self-assessed and true audit-readiness in food manufacturing
Food Industry Executive Report · 2026 Edition

Compliant
on Paper

Most food manufacturers think they're compliant with food safety regulations. Until they're not. This report measures the distance between what the paperwork shows and what auditors find when they walk the floor.

FDA inspection data · USDA FSIS investigations · Peer-reviewed meta-analyses · Certification body reports · Exclusive Yiannas interview

Published by Food Industry Executive · Presented by SafetyChain

Most food manufacturers believe they're compliant. Until they're not. An inspection finds mold around a handwashing sink. A third-party auditor raises a major non-conformance on sanitation. A recall traces back to a cleaning protocol that everyone was following on paper.

You may have checked all the boxes on the paperwork. But what will an auditor see when they walk the floor? How compliant will you prove to be when an inspector shows up, or when an outbreak occurs?

To quantify the gap, we reviewed FDA inspection data, USDA investigations, peer-reviewed meta-analyses, third-party certification body reports, and insights from Frank Yiannas, former FDA Deputy Commissioner for Food Policy and Response.

95% vs 50%
Self-reported vs observed hand-washing compliance.Meta-analysis of 31 food handler studies, MDPI.
73%
Increase in FDA warning letters, H2 2025 vs. H2 2024.Reed Smith analysis of FDA Compliance Dashboard.
69
Noncompliance reports at Boar's Head Jarratt, VA facility in the 12 months before the 2024 outbreak.CBS News, Boar's Head plant inspection findings.
$1.92B
Industry recall costs from label errors, 2024.New Food Magazine / Loftware analysis.

Lead Finding

Self-reported compliance runs nearly double what observers actually see.

Across every study in a peer-reviewed meta-analysis of 31 food handler training studies, food workers told researchers they followed safety protocols at far higher rates than researchers could document through direct observation.

In one underlying study, 95% of food handlers said they washed their hands as required. Observation caught 50%. In the same study, 82.5% said they kept their hair covered. Observation caught 17.5%.

Aggregated across all 31 studies, the pattern held. Training interventions produced an effect size of 0.80 on self-reported practices, but only 0.45 on observed practices. Training moves the needle on what workers say they do roughly twice as much as it moves what they actually do.

Self-reporting captures intention. Observation captures actual behavior.
SELF-REPORTED VS OBSERVED COMPLIANCE, % OF WORKERS 0 25 50 75 100 PERCENT Hand-washing Self-reported 95% Observed 50% 45-point gap Hair covering Self-reported 82.5% Observed 17.5% 65-point gap What workers say they do What observers see them do

Section 01 · Intent vs Behavior

Self-report captures intention. Observation captures behavior.

People are not lying on surveys. They're describing the version of themselves that shows up on good days. The version that remembers to wash their hands when the line is slow, not the version that skips it when product is backing up at the check weigher. The distance between intention and behavior, sustained over time, is where non-conformances start piling up.

“If you look at any catastrophic event, when the smartest men and women investigate those, what do they always say? It's a culture that allows these types of things to happen. Sure, there's hard science, but it's the soft stuff that's the hard stuff.”Frank Yiannas, former FDA Deputy Commissioner, in an exclusive interview with Food Industry Executive

If your internal compliance tracking relies on self-reported checks, you're measuring intention. A qualified employee signing off on a food safety check says the check happened. It does not say the surface was clean, the equipment was calibrated, or the root cause was addressed the last time this control failed. Those are different data points. Most systems don't capture them.

Certification frameworks are catching up to what the research has shown for over a decade. GFSI's Benchmarking Requirements v2024 embedded culture in food safety management programs across recognized schemes. SQF Edition 10 already made it mandatory. The March 2026 release of GFSI's updated culture position paper reinforced that food safety culture must be measurable, actionable, and continuously improved.


Section 02 · Audit Findings

All 5 of BRCGS's most-cited food safety non-conformances come from one section of the standard.

4,715
Non-conformances raised against a single clause in 2024-25.BRCGS Annual Report 2024-25.

Audit data tells the same story from a different angle. BRCGS (Brand Reputation Compliance Global Standards), one of the most widely adopted GFSI-benchmarked certification standards, completed more than 38,000 audits across its standards in 2024-25, generating over 240,000 corrective actions.

Within the BRCGS Food Safety Standard, every one of the top five most-cited non-conformances fell under a single section: Section 4, Site Standards. The single most-cited clause was 4.11.1, which covers the basic requirement that premises and equipment be maintained in a clean and hygienic condition. That one clause drove 4,715 non-conformances in 2024-25 alone. The next most-cited clause, covering equipment design and construction, drove 3,322.

Facilities are not failing audits on novel FSMA 204 provisions or cutting-edge allergen controls. They're failing on the basics: clean equipment, sound building fabric, working sanitation procedures. The same things manufacturers have been building food safety plans around for decades.

The most common non-conformances all come from Section 4 on site standards.
Clause
Description
NCs raised
4.11.1
The premises and equipment shall be maintained in a clean and hygienic condition.
4,715
4.6.2
The design and construction of equipment shall be based on risk, to prevent product contamination.
3,322
4.9.1.1
Processes shall be in place to manage the use, storage and handling of chemicals to prevent chemical contamination.
3,284
4.4.8
Doors (both internal and external) shall be maintained in good condition.
3,007
4.4.1
Walls shall be finished and maintained to prevent the accumulation of dirt, minimise condensation and mould growth, and facilitate cleaning.
2,933
The discipline

TRENDS, NOT INCIDENTS. A single finding is an incident. Three in a quarter at the same station is a pattern. Ten in a year across similar categories is a leading indicator of an outbreak. Assign someone with cross-facility visibility to read the aggregate.


Section 03 · Execution Gaps

Federal inspectors keep citing the same execution gaps, year after year.

Enforcement patterns for FSMA follow a similar pattern, according to an independent review of Form 483 observations and FDA warning letters. The most frequent citations include:

  • Hazard analyses built from templates without facility-specific detail
  • Preventive controls documented but not implemented
  • Process controls with critical limits that aren't scientifically validated
  • Sanitation procedures that specify frequency without specifying agents, concentrations, or verification
  • Monitoring records with missing signatures, incomplete fields, or unqualified personnel performing the work
  • Validation activities absent entirely from the food safety plan

Across each of these categories, the paper program exists, but the obligation-level execution doesn't. FDA inspectors are trained to find that gap. And more often than not, they find it.

“All of the outbreaks and recalls that we see in the news, how often do you know what really happened and what was done to prevent it? You almost never hear that, because the root cause isn't done. If you do a good root cause and follow the data, the data doesn't lie. It tells you why it happened and what you have to do to prevent it.”Frank Yiannas

Two forces keep manufacturers locked into repeat findings. Most food safety systems are built to pass audits, not to catch drift between audits, and the tools most plants use (like clipboards, checklists, and weekly QA walks) weren't designed to catch cumulative degradation. Then there's the human side. Nobody wants to be the person who escalates a maintenance issue that stops the line, or flags a problem that makes their shift look bad. So small drift gets tolerated until it becomes the multiple noncompliance reports that precede a fatal outbreak.

The discipline

COMPLACENCY IS DANGEROUS. Equipment cleanliness and sanitation monitoring lead BRCGS and FDA citation lists because they are the things humans get tired of checking. Wherever a repetitive manual check can be paired with an automated trigger, build the trigger.


Section 04 · Recall Trends

Recall counts are rising. The most dangerous category is rising fastest.

36.4%
growth in Class I recalls (reasonable probability of serious harm or death) since 2021.

The FDA and USDA announced 320 food recalls in 2025, up from 296 in 2024. Total U.S. food recalls have grown 21.4% since 2021, with Class I recalls (the category carrying a reasonable probability of serious harm or death) up 36.4% over that same period.

Class I recalls grew nearly twice as fast as the total U.S. food recall count between 2021 and 2025.
GROWTH IN U.S. FOOD RECALLS, 2021 → 2025 All U.S. food recalls FDA and USDA combined +21.4% Class I recalls Reasonable probability of serious harm or death +36.4% Class I recalls grew at roughly 1.7× the rate of the overall recall count.

Undeclared allergens caused 39% of 2025 recalls. Foodborne pathogens caused 34%. Listeria was responsible for 21 of 22 outbreak-associated deaths in 2025, the same pathogen that drove one of the industry's worst outbreaks in years.

Foreign material recalls more than doubled, from 42 events in 2024 to 89 in 2025, a pattern pointing less to isolated incidents than to systemic gaps in equipment maintenance and facility controls.


Case Study · Boar's Head

A century-old premium brand. Twelve months of documented warning signs. Ten people dead.

Boar's Head had been a trusted premium brand for over a century. The Jarratt, Virginia facility where the 2024 Listeria outbreak originated operated under USDA inspection jurisdiction. The company had documented food safety programs, sanitation standard operating procedures, and environmental monitoring. By every measure the word “compliant” traditionally covers, it was compliant.

Then 10 people died.

69
noncompliance reports issued at the Jarratt facility between January 2022 and the 2024 outbreak.

In January 2025, USDA's Food Safety and Inspection Service released its review of the outbreak, encompassing inspection records from January 2022 through 2024. Inspectors documented meat and fat residue on equipment after post-operation cleaning, condensation dripping from ceilings onto exposed ready-to-eat product, mold and mildew around handwashing sinks, insect infestations, cracks and broken flooring holding moisture, and a facility without barriers separating processing lines.

The outbreak killed 10 people and sickened 61 across 19 states. The recall covered more than 7 million pounds of deli meat. The Jarratt facility closed.

Frank Yiannas was brought in as interim chief food safety officer in the aftermath. Speaking to Food Industry Executive, he described a dynamic he has seen across dozens of outbreak investigations:

“Oftentimes in a crisis like that, when those companies are first contacted, they have a state of disbelief. They just don't want to believe that it could happen to them.”Frank Yiannas

He added, “I prefer working with organizations that place safety first over reputation.” The devastating implication is that many don't. A program that prioritizes reputation tolerates a lot of floor drift before it escalates. A program that prioritizes safety escalates sooner, and eats short-term pain to prevent long-term catastrophe.

The 69 noncompliance reports did not hide. They were documented, filed, and discussed. They accumulated because nothing in the system escalated their cumulative weight. The pattern was visible in the data the entire time. It did not become legible until 10 people were dead.


Section 05 · Enforcement

FDA warning letters jumped 73% in the second half of 2025.

In June 2025, the FDA launched an internal AI system called Elsa that flags high-risk facilities by analyzing historical observations, adverse event reports, and compliance data anomalies. The agency issued 327 warning letters between July and December 2025, a 73% increase over the same period in 2024. Untitled letters jumped from 5 in FY2024 to 58 in FY2025.

327
FDA warning letters issued July through December 2025. A 73% increase over the same period in 2024.
FDA enforcement volume climbed across categories in 2025.
FDA ENFORCEMENT GROWTH, 2024 → 2025 Warning letters H2 2024 → H2 2025 (July–December) +73% H2-2025 total: 327 letters Untitled letters FY2024 → FY2025 11.6× FY2024: 5 → FY2025: 58

Facilities that have coasted on routine certification cycles should not expect routine to continue. The current environment keeps people like Yiannas, who now advises companies as a consultant after leaving the FDA, busy.

“I have a hard time keeping up with the recalls and the outbreaks, because there's almost what I call a weekly, almost a daily drumbeat. As long as those events keep happening, there's work for us to do.”Frank Yiannas

Section 06 · Cost of Failure

Label errors alone drove $1.92 billion in industry recall costs in 2024.

A food recall is an expensive mistake, with nearly a quarter of recalls exceeding $30 million. Those numbers only cover costs like product retrieval, disposal, notification, and labor.

Then there are the indirect costs. Brand damage. Lost shelf space. Litigation. Retailer penalty clauses. Event studies of Class I meat and poultry recalls show an average loss of $109 million in shareholder value within five trading days of the announcement.

Label errors alone, a category most manufacturers would file as clerical, drove an estimated $1.92 billion in industry recall costs in 2024.

$1.92B
industry recall costs from label errors alone, 2024.

Section 07 · Consumer Response

More than one in five consumers will never buy your brand again after a recall.

21%
share of consumers who say they will never buy any brand from a manufacturer after a recall.

Harris Interactive research shows 55% of consumers will temporarily switch brands after a recall, 15% will never buy the recalled product again, and 21% will never buy any brand from the manufacturer again.

More than one in five consumers walk away from the entire manufacturer, not just the recalled product.
Share
Consumer response after a food recall
% of consumers
55%
Will temporarily switch to a competing brand.
55%
15%
Will never buy the recalled product again.
15%
21%
Will never buy any brand from the manufacturer again.
21%
Closing

What audit-ready looks like in practice

From the outside, a manufacturer that's closed the compliance gap looks identical to one that hasn't. Same walls, same PPE, same audit paperwork. The differences live in places most plant tours never see.

A supervisor flags a wet floor at 2 a.m. Escalating issues isn't punished, so escalation actually happens.

A QA manager doesn't sprint the two weeks before a certification audit. The same figures she pulls on a random Wednesday is the same data an auditor would see.

A plant manager reviews cumulative non-conformance trends alongside individual findings.

The CEO isn't the last to know. When a pattern starts forming at one plant, they can see signal the same week the plant manager does, not in a quarterly review.

These capabilities already exist. The plants that have built resilient food safety programs didn't invent a new compliance philosophy. They invested in the culture and systems that made their existing one visible and consistent. That's the work that defines the next decade of food manufacturing.

Five disciplines to close the gap

01

Measure what you actually do, not just what you say you do.

Self-reported compliance is a planning input, not an assurance. Every critical control should have at least one verification method that doesn't rely on the operator to self-report.

02

Read any noncompliance data as a trend, not a stack of incidents.

A single finding is an incident. Three in a quarter at the same station is a pattern. Ten in a year across similar categories could be a leading indicator of an outbreak. Assign someone with cross-facility and cross-function visibility (not line accountability) to analyze the aggregate.

03

Treat the boring findings as the dangerous ones.

Equipment cleanliness and sanitation monitoring lead BRCGS and FDA citation lists because they're the things humans get tired of checking. Wherever a person has to remember to check the same thing repeatedly, build something that checks it for them. A sensor. An alert. A scheduled task that doesn't require remembering.

04

Invest in culture as infrastructure.

Treat food safety culture as a requirement, not a trend. Survey the plant floor twice a year. Document what the survey turns up. Assign someone to close the gaps. When auditors ask what you did with the results, have an answer.

05

Compress the time between problem and response.

Boar's Head had 12 months of warning signs. Blue Bell had earlier signals on Listeria before its 2015 outbreak. Almost every major food safety failure of the last decade shows the same lag: weak signal detected, not escalated, catastrophe follows. Your facility already generates those signals. The question is whether your systems are designed to surface them early.

Methodology

Food Industry Executive produced this report in partnership with SafetyChain, the digital plant management platform built for food and beverage manufacturing and trusted by more than 2,500 facilities. The research draws on FDA inspection and recall data, USDA Food Safety and Inspection Service reports, peer-reviewed food safety research, certification body publications, and an exclusive Food Industry Executive interview with Frank Yiannas, former FDA Deputy Commissioner.

© 2026 Food Industry Executive Executive Edition · Presented by SafetyChain