Case file
- What happened: In July 2024, Boar's Head recalled more than seven million pounds of deli meats after CDC and USDA investigators linked a deadly Listeria monocytogenes outbreak to its Jarratt, Virginia plant.
- Scale: Roughly 60 hospitalizations and 10 deaths across about 19 states — the deadliest US listeriosis outbreak since 2011.
- Root cause: Years of tolerated sanitation nonconformities — mold, insects, rust and standing water documented by USDA inspectors but never driven to true root-cause elimination.
- The bill: Ten lives, a permanently closed facility, the end of liverwurst production, and brand damage no ledger will capture.
Here is an uncomfortable observation from two decades of chasing nonconformities across automotive and aerospace plants: the paperwork at Jarratt probably looked fine. Findings were recorded. Responses were filed. The loop closed — on paper — again and again. Ten people died anyway. Reading the publicly released USDA reports, I didn't see a plant that failed in silence. I saw a plant that failed loudly, in triplicate, for years, while everyone involved learned to live with the noise.
The situation
Boar's Head is not a discount brand. It trades on quality — the logo alone commands a premium at the deli counter. That is precisely what makes this file worth studying. Jarratt was an aging facility producing ready-to-eat deli meats, a category where Listeria monocytogenes is a known, well-characterised killer. Listeria doesn't need a novel failure mode to become lethal. It needs a wet niche, a biofilm, and a sanitation programme that has quietly shifted from elimination to coexistence.
The USDA inspection records released after the outbreak document conditions that would fail a walkthrough at any IATF 16949 plant I have ever audited: mold on food-contact and non-contact surfaces, insects, rust flaking into production areas, and standing water. Standing water — in a ready-to-eat facility, where moisture is the single most controllable listeria risk factor. None of this appeared for the first time in 2024. It appeared over years.
How it unfolded
The pattern will be familiar to anyone who has run a CAPA log. Inspectors found deficiencies. Reports were written. The plant responded — cleaned, patched, promised. The cycle reset. The next inspection found the same class of problem. Over time, "recurring finding" stopped being an alarm and became a category, an expected line item in the relationship between regulator and plant.
Then, in 2024, whole-genome sequencing did what paper never does: it tied the listeria in sick patients across roughly 19 states back to products from one plant. The recall exceeded seven million pounds. About 60 people were hospitalised; ten died. Boar's Head permanently closed Jarratt and discontinued liverwurst — an admission, in effect, that the facility could not be brought back.
Root-cause anatomy
Technically, this is textbook. Listeria is an environmental organism. It establishes harborage in exactly the places engineers can enumerate in advance: drains, dead legs in piping, worn gaskets, cracked floors, niches behind equipment where sanitation chemicals never reach. Once a biofilm forms, routine cleaning becomes containment — you reduce the population, you don't eliminate the niche. In PFMEA language, every one of those sites is a failure mode with a severity, occurrence and detection score before anyone eats a sandwich.
Organizationally, it is worse. A plant that decays behind a strong brand is not a sanitation problem; it is a governance problem. Somewhere up the chain, a decision structure existed in which an aging plant's condition was known, tolerated and budgeted around. Nobody signs a document approving mold. But capital allocation is a document, and year after year the money that would have fixed floors, drains and equipment went elsewhere. Tolerated findings are a strategy, even if no one calls them one.
When the same finding appears for the third time, you no longer have a nonconformity — you have a policy.
Where the quality system failed
Name the discipline and it failed here. On CAPA governance: a corrective action that closes on "cleaned the area" is containment (8D steps D1–D3), not correction (D4–D6). When a finding recurs, the honest conclusion is that root cause was never found — the team treated the symptom's location as its cause. On environmental monitoring: swabbing for listeria is risk analysis, not housekeeping. A trending positive near a drain is a leading indicator of an outbreak; treated as a cleaning chore, it is noise. On audit programme design: an audit that records the same deficiency annually without escalating severity has become a ritual, not a control. And on change control of the asset itself: equipment aging, floor degradation and layout changes alter the risk picture continuously. If your PFMEA was written for the plant as commissioned, it describes a building that no longer exists.
In aerospace we live with a rule that translates directly: any repeat finding from a prior audit is automatically an escalation, because recurrence proves the previous corrective action failed. A system without that automatic escalation will, given enough years, produce Jarratt.
What would have caught it
- A PFMEA for the environment, not just the process. Score drains, dead legs, standing-water zones and biofilm niches as failure modes with severity, occurrence and detection. Anything above the action threshold gets engineered out — regraded floors, redesigned drains, replaceable gaskets — not cleaned more often.
- A recurrence gate on every CAPA. Any nonconformity matching a prior finding within 24 months goes to senior management with a mandatory root-cause re-analysis and a "why did the last action fail?" line item. No exceptions, no housekeeping classifications.
- Environmental monitoring trended as risk data. Routine swabbing of harborage sites, escalation on any trend toward positives — reviewed monthly by someone with authority to stop the line and spend money.
- Governance of aging assets. An annual plant-condition review that forces a funded decision: refurbish, contain or close. "Defer" must appear in the minutes as what it is — an accepted risk with a named owner.
My take
Across automotive and aerospace, I have watched audit programmes decay into paperwork more than once. I have sat in review meetings where a repeat finding from last quarter was discussed with the same calm as the weather. At Witte Automotive I learned the hard way that closing a quality loop means the finding does not come back — and that the only honest metric for a CAPA system is recurrence rate, not closure rate. A closed finding that returns was never closed. Building a QA organisation from scratch for 900+ people at SNOP taught me the other half: systems drift because nobody owns the drift. You build the escalation before you need it.
What strikes me most about Jarratt is how much warning there was. In my world we measure the cost of poor quality and work it down — I've driven defect costs down 70% with QRQC and disciplined A3 thinking. But CoPQ assumes you survive the failure. Ten deaths are not a cost line. The sanitation crew, the drain redesign, the plant refurbishment — all of it was cheaper than one week of what followed, and everyone with authority knew the conditions existed. They just never felt them.
What this means on your floor
- Any finding that repeats is proof the last corrective action failed. Escalate automatically; never re-close with the same containment.
- Environmental conditions are failure modes. Score them in your PFMEA and engineer out the niches you can predict today.
- Recurrence rate, not closure rate, is the real health metric of your CAPA system.
- Deferred maintenance on a food or safety-critical asset is an accepted risk. Make someone senior sign it, by name, every year.
The Jarratt plant did not fail in silence. It failed with documentation. The mold, the insects, the standing water — recorded, answered, back the following year. That is the lesson I would pin to every CAPA board in every plant I run: a finding written down is not a risk controlled. Only elimination counts, and the loop either closes at root cause or it eventually closes on you.