Food Manufacturing & Processing

The stake: FSMA preventive controls and third-party audit schemes that require documented pest programs.

What is genuinely at stake in a food plant?

A supply relationship, a regulatory position, and in the worst case a product decision — with the pest itself appearing fourth on that list.

Manufacturers, processors and packers work under a framework that changed the question being asked. The old question was whether contamination had occurred. The current one is whether the plant can demonstrate that it analyzed its hazards, controlled them, monitored the controls, acted when monitoring said something was wrong, and verified the whole arrangement works. That is a documentary standard as much as a physical one, and it is why a plant with no visible pest activity can still be in trouble, while a plant with a documented receiving-door problem being driven down week by week is in reasonably good shape.

Above the regulation sits the audit layer, which in most plants applies more weekly pressure than any regulator. A customer specifies a certification scheme and a minimum outcome as a condition of supply, and the pest element is scored inside it. A downgrade costs a listing, or an early customer audit, and none of that requires anything ever to have been in the product.

At the far end sits the event nobody plans for. Insect fragments in an ingredient stream, rodent contamination in storage, a bird in a packaging hall — each is a product disposition decision before it is a pest decision, and each is capable of becoming a recall.

So the operator’s question is never whether there are insects on the site. There are — at the perimeter, in the yard, on the inbound pallets. The question is whether the plant can show where they are, what is being done, and which way the trend is going.

What does the preventive-controls framework actually require of the building?

More than most plants realize, and in unusually blunt language.

The good manufacturing practice requirements for sanitary operations are not hedged. Pests “must not be allowed in any area of a food plant,” and effective measures must be taken to exclude them from the manufacturing, processing, packing and holding areas. Pesticide use inside the plant is permitted “only under precautions and restrictions that will protect against the contamination of food, food-contact surfaces, and food-packaging materials.” Two sentences, and between them they set the whole strategy: exclusion is the obligation, and chemistry is the constrained exception.

The definition of the thing being excluded is broader than the word suggests. In the rule, a pest is “any objectionable animals or insects including birds, rodents, flies, and larvae.” Birds are named first among the examples, which is worth noticing in a plant where nobody has been on the roof since the refrigeration project.

The grounds requirements are equally specific and are the most commonly under-read part of the rule. Grounds must be kept in a condition that protects against contamination, which includes properly storing equipment, removing litter and waste, and cutting weeds or grass in the immediate vicinity of the plant where those “may constitute an attractant, breeding place, or harborage for pests.” It includes maintaining roads, yards and parking lots so they are not a source of contamination, and adequately draining areas that could provide a breeding place. It requires screening or other protection against pests where necessary, and it addresses the case where the adjoining grounds are not under the operator’s control — the plant is still expected to take adequate precautions to exclude what comes from next door.

Read together, those are structural instructions. A perimeter strip that is hard, clear, drained and swept satisfies several of them at once and removes a substantial share of rodent harborage without anything being applied. The same paragraph explains why an auditor spends the first twenty minutes of a plant walk outside the building.

Pest management is normally handled as a prerequisite program rather than as a preventive control in its own right, though a sanitation control can be written to include elements of it where the hazard analysis justifies that. The placement is sometimes read as a demotion. It is the opposite: a prerequisite is the ground the hazard analysis is allowed to stand on, so a pest program that exists on paper only quietly invalidates the reasoning above it.

What does a PCQI actually sign off on?

Four things, none of which a pest contractor performs, and all four of which are made easier or harder by the state of the pest records.

A preventive controls qualified individual is defined as a qualified individual who has completed training in the development and application of risk-based preventive controls at least equivalent to a standardized curriculum recognized as adequate by FDA, “or is otherwise qualified through job experience to develop and apply a food safety system.” The rule then names what must be done or overseen by that person: preparation of the food safety plan, validation of the preventive controls, oversight of the review of records, and reanalysis of the food safety plan. The training itself has to be documented, including the date, the type of training and the person trained.

Each of those touches the pest file at a specific point.

Preparation of the plan is where pest management is placed — as a prerequisite, or inside a sanitation control — and where the plant states what the program is expected to deliver. A plan that describes a program the site does not actually run is a finding waiting for someone with the patience to compare them.

Validation asks whether the control does what it claims. For a sanitation control with a pest element, that is where the evidence has to come from somewhere other than the contractor’s assurance.

Records review has a clock on it: the rule contemplates review within a set period and requires written justification if it takes longer. Pest service records that arrive late, in a format nobody can read, or with the findings not separated from the attendance are therefore an obstacle to a signed-off review rather than a minor irritation.

Reanalysis is triggered by change — a new process, a new ingredient, a new building, a significant new hazard. A plant that logs its projects and equipment moves can explain a pest spike during reanalysis. A plant that does not will be explaining it as a program failure.

Separately, corrective actions and corrections have their own requirements: procedures must identify and correct the problem, reduce the likelihood of recurrence, evaluate the affected food and prevent adulterated product entering commerce, and all corrective actions taken must be documented in records. A correction is the lighter instrument, available for a minor, isolated problem addressed promptly. Knowing which of the two a given pest finding is — and being able to defend that classification — is a QA decision that depends entirely on the quality of what the pest contractor wrote down.

Ryan Katz holds PCQI and HACCP, both issued by AIB International, and is SQF credentialed — which is why these conversations happen in the plant’s own vocabulary rather than in a pest contractor’s. The PCQI certificate matters more here than it looks: under the preventive controls rule the qualified individual who develops a food safety plan need not be an employee of the facility, which is what allows the pest sections of a plan to be written by someone who understands both halves. Graduate is also a member of the National Pest Management Association and the New York Pest Management Association, which are trade memberships rather than credentials and are not presented here as anything else.

One honesty note, because credential inflation is endemic in this trade. Those certifications belong to a person rather than to a company, and Ryan Katz holds no entomology degree and no entomological certification. The entomologist here is his father, Arnold Katz, who holds the B.S. in Entomology from the University of Georgia and still provides identification and diagnostic support where a species determination changes the plan. What each credential does and does not permit is set out on Ryan Katz’s page.

What does each audit scheme expect the pest program to produce?

Broadly the same six artefacts, with meaningful differences in emphasis and in how explicit each scheme is about the details.

SQF. Administered by the SQF Institute and recognized by the Global Food Safety Initiative. Its own pest prevention guidance expects the program to outline the frequency with which pest status is checked and to identify the location of bait stations, traps and chemical sites, covering product, material and packaging storage as well as the main processing areas. It expects records to include service reports, pesticide usage logs, pest sighting logs, corrective action reports and trending of activity — trending named alongside the rest rather than left implicit. It expects the site itself to record sightings and frequency rather than leaving that entirely to the contractor, contamination events to be investigated with the disposal and outcome documented, external operators to be licensed by the relevant authority, and chemicals to be approved, stored separately from food materials and supported by accessible safety data sheets. What happens when an auditor meets a program older than the standard is covered on our food safety and SQF programs page.

BRCGS. The Global Standard Food Safety, published by BRCGS and also GFSI-recognized, sets out safety, quality and operational criteria for a manufacturing site and handles pest management within its site standards section — clause 4.14 in Issue 9. The practical difference a QA manager notices is the standard’s insistence on demonstrated competence and on the site owning the outcome rather than delegating it to the contractor’s visit report.

PrimusGFS. Administered by Azzule Systems, GFSI-recognized, and the scheme most often met where the operation touches produce — cooling and cold storage, packinghouse, processing and distribution all sit under its GMP audits, with pest control assessed in the facility module.

AIB International. Not a GFSI scheme but long-established across North American baking and processing, and unusually explicit. Its consolidated standards treat integrated pest management as one of five categories — alongside operational methods and personnel practices, maintenance for food safety, cleaning practices, and the adequacy of prerequisite and food safety programs — and define it as the assessment, monitoring and management of pest activity to identify, prevent and eliminate the conditions that could promote or sustain a pest population. Scoring runs in bands from no issues observed down to unsatisfactory, with the category score set by the most significant observation and adjusted downward for multiple findings, which is why a single serious structural condition can drag a category regardless of how tidy the rest of it is. The standard requires a written IPM program that incorporates the site’s other prerequisite programs; a copy of the certification or registration document for each person performing pest management services; a current and accurate site map listing the locations of all pest-monitoring devices; service records in the devices that match the documentation held in the facility; a quarterly review of the log by pest management personnel to identify trends, with a report of findings submitted to designated facility staff; exterior bait stations that are tamper resistant, positioned, anchored, locked and labeled; and — the clause that surprises people — toxic and non-toxic commercial baits not used for interior monitoring at all.

The common denominator across all four is worth stating, because it is what a program should be built to satisfy regardless of which scheme the customer specifies: written scope, accurate map, executed schedule, read data, closed corrective actions, verified competence. A program built for those six is portable between schemes. A program built for one scheme’s clause numbering is not.

How should the device network be mapped and numbered?

So that the map still describes the building eighteen months after a warehouse reorganisation, which is a discipline problem rather than a drawing problem.

Start with what the map is for. It is the document an auditor uses to decide whether to sample the device network or verify all of it, and the fastest way to lose that argument is a station that is not where the drawing puts it. So the map records every device with a permanent identifier and a located position, carries a revision date, and is redrawn the week the plant changes rather than the week before an audit.

Numbering by location logic, not installation order. Numbering in the sequence somebody happened to install becomes unreadable the first time three devices are added and two removed. Numbering that encodes the area and runs in a consistent direction — exterior stations clockwise from a fixed datum, interior devices by zone then by wall run — survives change, because a new device gets a number that says where it is rather than when it arrived.

Density follows risk, not geometry. Devices concentrate inside every exterior opening, at receiving, along interior perimeter walls, at transitions between hygiene zones and around ingredient storage, and thin out where there is nothing to intercept. Perfectly even spacing is the clearest sign a layout was transcribed from another site.

Each device class answers a different question. Exterior stations describe the pressure arriving at the building and the harborage feeding it. Mechanical multi-catch devices immediately inside doors are diagnostic: repeated captures inside one opening is a statement about that door, and the answer is a seal or a leveller repair rather than another device. Interior monitoring devices along wall runs describe whether anything is moving through the building. Light traps describe flying pressure and, more usefully, its source — a catch tray identified to species separates a receiving-door problem from a drain problem from something that arrived on a pallet. Pheromone traps in storage describe stored-product populations that are otherwise invisible until they are in the product.

Light trap placement has rules that are routinely broken. Units are positioned to intercept rather than to attract: set back from doors, out of any sightline from outside the building, and never above exposed product or a food-contact surface. A trap mounted where it is visible through an open dock door is drawing insects across the threshold it is meant to defend.

A pheromone grid is a locator, not just a counter. Species-specific lures for the flour and grain beetles, warehouse and cigarette beetles and the Indianmeal moth complex, laid out on a grid through storage, will point at a source if the counts are plotted spatially rather than totalled. Three traps in a line with a rising gradient towards one bay is a harborage in that bay — a split sack behind a rack leg, spillage under a pallet, an old lot that never rotated. That is a level of resolution no amount of general service produces.

Every device is checked on schedule and the result recorded whether or not anything was in it. A blank line and a recorded zero look identical in a binder and are not the same claim, and only one of them is defensible.

What does trend analysis actually look like when it is done properly?

Four things at once, and a total is none of them.

Two plants can post identical monthly numbers and be in entirely different conditions. Spread across a large network, the number is background pressure. Concentrated in four devices at one end of a building, the same number is a structural statement about one opening that will reproduce itself every month until somebody replaces a seal. Reporting a site total conceals that difference by design.

Plot by device and by area. Pressure needs a location on a floor plan before it can be acted on. The unit of analysis is the device, aggregated to the zone, not the site.

Plot over enough time to separate season from structure. Autumn rodent movement indoors, spring and summer fly pressure, and a stored-product insect population whose generation time shortens as the warehouse warms are all seasonal and all expected. A penetration that has been open since a refrigeration project is not. The two look similar over eight weeks and completely different over two years, which is the argument for keeping historical data in a form somebody can actually query.

Mark interventions on the timeline. A corrective action is judged by what the counts did afterwards, not by the fact that somebody signed it off. A seal replaced in week nine should be visible in the device data by week twelve; if it is not, the diagnosis was wrong and the loop needs to run again rather than close.

Set action thresholds in advance and in writing. A threshold decided after the fact is a rationalisation. Thresholds differ by device class, by zone and by species — one Indianmeal moth in a raw storage pheromone trap means something different from one in a packaging hall — and they should specify what happens when they are crossed, including who is told and within what period.

Then two anomaly rules, applied in both directions. Sustained catches with no corrective action attached mean the plant is measuring a problem and declining to act on it, which is the single worst pattern a file can show. Sustained zeros in a device that ought to see something mean the device is obstructed, damaged, unreachable or unchecked, and each of those is worth verifying physically rather than assuming.

The test of a reporting package is whether a food safety director can answer three questions from it inside a minute: where is the pressure, which way is it moving, and what did we change. A package that cannot is a filing system.

What does the corrective-action loop look like on the floor?

Containment, correction, verification, then the question almost every log forgets to ask.

Take a real sequence. Monitoring shows repeated captures in the mechanical device inside bay four. Adding devices there is containment — it improves the measurement of a problem that is still happening, and it is often the right first move, but it is not correction and should never be logged as though it were.

Correction begins with root cause, and in a plant the root cause is usually several things at once rather than one. For bay four it might be a compressed bottom seal, an open back corner in the leveller pit, and a habit of leaving the bay open through a forty-minute unload. Each of those gets an owner and a date: the seal to maintenance, the pit to the contractor or a fabricator, the unload practice to the shift supervisor. Verification is what that device does over the following weeks, recorded and compared against what it was doing before, with evidence attached: a photograph of the closed pit, a follow-up count, an inspection note. A closure logged without any of those is a claim rather than a fact.

The forgotten question is the preventive one. If bay four’s pit is open at the back, what condition are the other five in? Were all the seals fitted in the same year and therefore near the same failure point? Is the unload habit shift-specific or plant-wide? A register that only ever fixes the device that flagged works through a building one failure at a time, indefinitely, and anyone reading a year of it will say so.

Three practical rules make a register converge rather than accumulate. Most items do not belong to the pest contractor — they belong to maintenance, engineering, sanitation or purchasing, and an item without an owner inside the plant does not close. A target date that has been quietly re-dated three times is worse in the file than an item honestly recorded as waiting on a capital cycle with the justification written beside it. And a recurring finding should trigger a review of the category rather than another repeat of the same entry, because the second identical finding is information about the system and the fifth is information about the register.

Where do sanitation and pest management actually meet?

At a small number of interfaces that generate most of the persistent problems in most plants, and where each department reasonably believes the item belongs to the other.

The master sanitation schedule is the meeting point on paper. Daily cleaning covers what production sees; the periodic schedule covers what it does not — behind and beneath equipment, inside voids, above the line, the corners no sweeper reaches, the drains under the platforms. Almost every chronic pest population in a food plant lives in something on the periodic schedule that keeps slipping, and the fastest diagnostic available to a new contractor is to read that schedule and ask which items have moved twice.

Drains are the clearest case. Biofilm in a floor or trench drain will support a small fly population entirely independently of anything else happening in the building, and no amount of exterior work touches it. The correction is mechanical cleaning of the drain wall and the underside of the cover, on a schedule, verified — not a product poured down it.

Condensate and standing water. Uninsulated lines dripping onto a floor, an overflowing chiller pan, a low spot in a slab holding water after every wash-down. Water availability decides whether a cockroach or fly population establishes at all, and it is usually a maintenance item logged as housekeeping.

Wet cleaning in a dry plant. Where the process is dry, introducing water is a hazard in itself and sanitation is generally dry — vacuum, scrape, brush. That constrains what can be done about spillage inside a void and makes equipment design a pest question. It also means any controlled deviation to wet cleaning should reach the pest program in advance, because a wetted-out area behaves differently for weeks.

Equipment voids and hollow framing. Open-ended frame members, motor housings, conveyor undersides, guarding that cannot be removed without a fitter. Simultaneously a cleaning failure and a harborage, which is exactly why they survive so long: neither department can close one alone.

Spillage behind and beneath racking. The most reliable source of a warehouse stored-product population, and the one a pheromone grid locates.

The ownership split should be settled in writing when the account starts rather than during an audit week. Naming the condition, running the devices, holding the data and writing the report are ours. The drain schedule, void access, storage practice and the physical repair are the plant’s. Written down and reviewed on a cycle, those findings close.

Where does a plant actually leak, structurally?

Three places dominate, and they are the three that a house-scale idea of exclusion does not prepare anybody for.

Dock levellers. The single highest-yield defect in almost every plant. A pit-style leveller sits in a recess below the dock floor, and the recess is warm, sheltered, collects debris and organic material, and connects directly to the interior at the back and sides. The lip and the header seal against the trailer, not against the building, so with no trailer at the bay the pit is an open room with a hole in the wall. What fails: the bottom weather seal compresses and cracks; the side curtains tear on the first mis-spotted trailer; the back corners of the pit are open by construction and were never closed; the pit fills with dunnage and sweepings nobody removes because nobody owns it. The correction is a proper pit enclosure and seal detail, brush or compression seals maintained on schedule rather than replaced after a finding, levellers written into the periodic sanitation schedule, and a sealed pit system when the equipment is next replaced. The operational half matters as much: a bay left open through a long unload has defeated every one of those details for the duration.

Utility penetrations. The most numerous defect in any plant by a wide margin, and the most under-inspected. Every trade that has ever worked in the building cut something and very few closed what they cut: conduit and pipe sleeves through walls and slabs, refrigerant and glycol lines, compressed air, steam and condensate returns, drain lines through floor slabs, data and low-voltage runs added a decade after the electricians left, and roof curbs with open corners above the ceiling grid. The annulus around a pipe is the classic case — a two-inch sleeve carrying a one-inch pipe is a permanent quarter-inch-plus gap and a mouse needs a quarter-inch. The correction is a packed and sealed annulus: copper or stainless mesh packed into the void, then cement or mortar at masonry, or custom-fabricated 26-gauge galvanised sheet metal collars mechanically fastened where the substrate will not take mortar. What it is not is expanding foam. Foam is the industry’s most common failure here — rodents chew straight through cured foam, and in an audited plant a foamed penetration is a finding written twice: once as an open penetration and once as a repair that misrepresented itself as one. We do not use it in any role.

Rail doors. In a rail-served plant these are the largest openings in the building and the least maintained, because they operate a handful of times a week and nobody thinks about them in between. The track slot passes under the door and through the wall, which is a permanent opening the door cannot close, and the rail and ballast make a flat, sealable threshold impossible. Cars then stand at the opening for hours with the door open. The realistic corrections are a hardened, cleanable rail slot detail; an interior secondary barrier treated as the real seal, with the rail door as weather protection; brush seals at the jambs maintained on schedule; and housekeeping for the ballast and apron outside, which is otherwise ideal harborage sitting against the building line.

Around those three sit the familiar ones: personnel door sweeps and closers, air curtains that need specifying, aiming, balancing, switching on and periodically verifying, screening on vents and intakes, drain cover and trap integrity, wall-to-floor junction repair, and the roof — under-inspected in most plants and frequently where the real answer is.

Most of this is capital work carried out by other trades, which is why a finding has to be written specifically enough to be scoped, priced and then verified. A note reading “seal the dock area” will be actioned by nobody. The principles are on structural exclusion, and where a plant is building or gutting a line, the cheap version of this conversation happens before the walls close — which is what exclusion consulting is for.

What documentation package should a plant hold?

Eight items, current, in one place, legible to somebody who has never been in the building.

The written program — scope, pests of concern for this facility, frequency and its justification, device types and the reasoning behind their placement, action thresholds, escalation, and who is responsible for what on both sides. The site map, current, with a revision date. Service reports that name findings by device and area rather than recording attendance. Material records for anything applied: what, where, when, at what rate, under which label, with the label and safety data sheet retrievable — a legal record before it is an audit artefact. Trend data in a form that shows direction without anyone reading a year of notes. The corrective-action register, with root causes, owners, dates, closure evidence and preventive actions. Contractor documentation — licensing, insurance and training records, verified by the facility rather than assumed. And change records: construction projects, equipment moves, new suppliers, new raw materials.

That last one earns its place, and two other failures belong beside it. Construction is the strongest single predictor of a new pest problem in a stable plant: a site that logs its projects can explain a spike, and a site that does not will have the spike read as program failure. Stored-product insects arrive in ingredients, so a plant with an excellent perimeter and no incoming inspection keeps receiving populations by the pallet — plot the pheromone data against delivery dates and it is usually obvious in an afternoon. And data can be collected dutifully and never read, with the pressure visible in the numbers for a year before the auditor becomes the person who notices it.

Where does a food plant program start?

With the last audit report, the findings carried forward more than once, and a walk of the building.

Graduate has worked on buildings across New York City and Long Island since 1983, on the stated position that pest problems are building problems needing permanent structural solutions. In a food plant that stops being a philosophy and becomes an operating constraint, because the building is most of what a plant is permitted to change. This is also not a recent specialism here: the food-safety pest program this firm built for a manufacturing client in the mid-1990s turned out to be what SQF later required, unchanged, and that account still runs. The industry gave the practice a name about fifteen years ago.

The scheme relationship and the questions worth asking any contractor before an audit are on food safety and SQF programs, and the wider framework is on commercial pest management. Where part of an operation is foodservice or retail, the inspection regime differs and sits under restaurant pest control; multi-site operators will recognize the consistency problem under property management. Technically, the exterior and below-grade rodent picture is on rodent control, the drain-associated pressure behind persistent small fly problems under cockroach control, and identification in the pest library. Coverage is listed under locations.

To begin, contact us with the facility type, the scheme you are audited against, when the next audit falls and what the current data is showing. The consultation costs nothing and is usually a phone call. A written proposal and plan carries a service fee, because the specification is the deliverable, and that fee is credited against the project if the work goes ahead.