Most environmental monitoring programs I see are run in a clean, cold, empty plant. Sanitation finishes at 4 a.m., the swabs go down at 5, the results come back negative, and the report gets filed. The program is compliant. It’s also, in the way that matters, telling you almost nothing.

A preoperational swab answers one question well: Did the cleaning crew do the job they were assigned last night? That’s a useful question, but it’s not the question an environmental monitoring program exists to answer. The question that matters is whether something is living in this building that will reach product while the line is running. A plant that has just been cleaned and sanitized is the worst possible time to look for it.

Two questions, one program

The regulation is less prescriptive than people assume, and more demanding. Under 21 CFR 117.165(a)(3), environmental monitoring is required as a verification activity where contamination of a ready-to-eat food with an environmental pathogen is a hazard requiring a preventive control. The part worth reading twice is 117.165(b)(3): The written procedures must be scientifically valid, and the timing and frequency of sample collection and testing must be adequate to determine whether preventive controls are effective.

“Adequate to determine whether preventive controls are effective” is a performance standard, not a schedule. A program that only ever samples a sanitized, idle plant cannot meet it, because it has no opportunity to detect the thing it’s looking for. There’s a laboratory problem on top of the sampling-design problem: Swab too soon after sanitizing and, in the U.S. Food and Drug Administration’s words, “the sanitizer may not be adequately neutralized and could interfere with the analytical test” (Section XIII.D.4). The preop swab is not only asking a narrow question; it’s asking it under conditions that can suppress recovery.

The practical fix is to stop treating environmental monitoring as one activity and start treating it as two with different designs:

Cleaning verification: Sampled immediately before and after cleaning and sanitizing, before operations start. This tells you whether a procedure was executed.

Harborage detection: Sampled during production. A 2022 review of environmental monitoring practice in Foods recommends sampling at least three hours after startup, or at the end of production before cleaning begins, precisely because that’s when a resident organism in a niche has had time, moisture, and traffic to express itself.

Both belong in the program. Only one of them finds the problem that causes recalls, and it’s usually the one that isn’t scheduled.

Why 3 hours in

The FDA says the same thing in its own voice. The draft guidance on Listeria monocytogenes in ready-to-eat foods states that “the most important time to collect environmental samples is at a time that is several hours into production (e.g., 3 to 4 hours)” or “preferably just prior to cleanup” (Section XIII.D.4). That isn’t a tolerance of production sampling; it’s a preference for it.

The reasons are physical. Condensate has had time to form and drip. Equipment has warmed, vibrated, and flexed, so the hollow roller, the cracked gasket, and the worn hinge are doing what they do under load rather than at rest. Forklifts have crossed from the raw side. Floor water has been moved by foot traffic. None of that exists at 5 a.m. on a dry, still floor.

The same review notes that resident strains have been documented persisting in facilities for periods ranging from seven months to seven years. Organisms that survive that long are not surviving on open stainless steel. They are living in places your sanitation chemistry never reaches, and they emerge under production conditions. If your sampling plan only ever visits the plant in its cleanest state, a seven-year resident can stay invisible for seven years while your trend chart stays flat and green.

Zone 3 positives are data, not failures

The zone framework is widely adopted and widely misread. Zones 1 and 2 are food-contact and adjacent surfaces; zones 3 and 4 are nonfood-contact surfaces within and outside the production area. Most programs weight their sampling toward zones 1 and 2, which is correct on risk. The misreading is cultural: A zone 3 positive gets treated as a bad grade.

It is the opposite. A positive on a drain, a wall-floor junction, or a wheel is an early warning collected from a place where no product is at risk. It’s the cheapest information the program will ever generate. The review makes the point plainly: Positives in zones 3 and 4 indicate harborage sites and early contamination, giving you the chance to act before product is affected.

If your organization responds to Zone 3 positives with blame, your sampling team will learn—quickly, and without being told—to swab the places that come back clean. At that point you are paying for a laboratory to confirm your floor plan.

But what about Zone 1?

This is the first question any quality manager asks, and an article that recommends production sampling owes an answer rather than a dodge.

The FDA’s recommendation isn’t confined to nonfood-contact surfaces. The draft guidance says plainly: “We recommend that an establishment test both FCS and non-FCS sites at each sampling time” (Section XIII.D.3). Read alongside the timing recommendation above, that means food-contact surfaces during production, several hours in.

The reason most plants don’t do it isn’t ignorance. A food-contact positive during production implicates the product that ran across that surface, and nobody wants to generate that finding on purpose. The predictable result is a program that samples where it’s safe to find nothing.

That cost is real. It’s also manageable, and the way it’s managed is by deciding everything in advance.

Test for the indicator, not the pathogen: Most programs test food-contact surfaces for Listeria spp. rather than for L. monocytogenes directly. The FDA is clear about what that buys you: A Listeria spp. positive “indicates the potential for contamination of an FCS or non-FCS with L. monocytogenes” and “does not establish the presence of L. monocytogenes” (Section XIII.D.2). The first result triggers investigation and confirmation. It’s not a verdict on the product.

Define the implicated lot before you swab, not after: The lot is whatever ran on that line from the last verified clean point to the time of sampling. Written down in advance, it’s a bounded, known quantity. Discovered in the moment, it’s an argument.

Have the corrective action written before you need it: The FDA’s guidance sets out corrective-action procedures for Listeria spp. and L. monocytogenes findings, including hold-and-test procedures for ready-to-eat food (Section XIII.F.3), and it directs that sampling frequency be increased when positives are found in the plant (Section XIII.D.4). A plant that has adopted those before it starts food-contact sampling during production is making a planned decision. A plant that hasn’t is improvising during a hold, which is where bad outcomes come from.

Then vector out rather than reswabbing: Expand sampling around the positive to find the harborage point, instead of recleaning the single site and retesting it clean.

The honest summary is that during-production, Zone 1 sampling costs you the option of not knowing. Plants that have made that trade generally find the exposure is bounded and the information is worth it. Plants that haven’t are usually carrying the same risk without the data. But it’s a decision a quality organization must take deliberately, with the hold procedure and the disposition criteria agreed upon in writing first, and ideally with the recall and legal side in the room when it’s taken.

Designing a plan that can actually find something

Four decisions do most of the work.

Site count: The Foods review summarizing the FDA’s recommendations notes that even small ready-to-eat facilities should sample at least five food-contact and five nonfood-contact sites per production line. Programs built around a fixed handful of easy, repeatable sites are measuring those sites, not the plant.

Rotation: Rotate sampling across shifts and days. A program that always samples Tuesday day shift has characterized Tuesday day shift.

Targeting: Go where water, residue, and movement collect—drip points, standing water, wall crevices, equipment hinges, high-traffic paths. Hard-to-reach is the point, not an obstacle.

Vector response: When a positive comes back, expand outward from it rather than simply reswabbing the same spot after recleaning. A clean reswab of a single site proves you cleaned a single site.

Trend, don’t grade

The most common structural flaw is treating each result as pass or fail. A single negative is weak evidence; a percent-positive rate by zone, tracked over time against a baseline, is a control chart, and it behaves like one. Shifts after a line change, a construction project, a staffing change, or a chemical switch show up in the rate long before they show up as a single alarming result.

This is the point at which environmental monitoring stops being a food safety chore and starts being a quality system input. Baseline, rotate, trend, investigate shifts, verify the correction with expanded sampling, and feed the finding back into the preventive control. That loop is ordinary quality practice. It’s also literally what 117.165(b)(3) asks for.

The test

Here is a quick diagnostic for your own program. Pull the last 12 months of environmental results and ask three questions: What proportion were collected while the line was running; how many distinct sites appear in the record; and what happened, procedurally, the last time one came back positive?

If the answers are “almost none,” “about a dozen,” and “we recleaned and reswabbed,” the program is working exactly as designed. The design is the problem, and fixing it takes decisions, not capital.