How the certification audit actually runs — the records review, the floor walk, how findings are graded, and the handful of gaps that write up most first-timers — so the day is a confirmation, not a scramble.
Do these first
Real help from real food people
Chef Diego runs a real food plant. If this page didn't get you there, tell us — a person reads every message.
After this lesson you can walk into your first certification audit knowing how the day runs — and having already found and closed the handful of gaps that fill up most first-timers' reports. The audit is more predictable than the dread around it. It asks two questions, in two parts, and it grades what it finds on a scale you can plan against.
Strip the audit down and the auditor is checking two things. First, does your written system meet the code? Second, does your floor actually run the way the writing says? The whole visit is those two questions, and the code has a plain shorthand for them: say what you do, do what you say. A perfect binder that describes a process the floor abandoned fails just as surely as a floor that runs well with nothing written down.
Those two questions map onto two parts of the audit. The first is a review of your records — your food-safety plans, your procedures, your training and monitoring records, your internal audits and corrective actions, and your traceability and mock-recall exercise. People call this part the desk audit. The second is the floor walk: the auditor observes the work and the cleaning, follows up on anything the records left unclear, and confirms the system actually operates across every product and process in your scope. People call that the facility, or site, audit.
Under , that records review may be done remotely by agreement, but at least half of the audit's allocated time has to be on-site, and the certification or re-certification audit runs a minimum of two days. The exact split and the names differ by scheme — some run the records review as a separate, earlier stage — so ask your certification body how yours is structured. What does not change is the pairing: paper against the code, then floor against the paper.
The floor walk covers the whole site, not just your scope
The on-site part takes in the entire site — interior and exterior — and every operational and cleaning shift, including the pre-operational inspections, regardless of which products are in your scope of certification. An auditor who arrives for a day shift can still ask about how the night shift cleans. Plan for the whole building, not the part you rehearsed.
The visit is evidence-gathering, not interrogation. The auditor builds their picture three ways: reading your documents and records, interviewing the people who do the work, and watching the operation and the cleaning happen. They are matching what they read to what they see, and writing down wherever the two diverge.
A few things make the day less mysterious. The auditor expects a real history to read: SQF recommends at least 90 days of records be available before the site audit, because a system with a week of records has not been running long enough to prove anything. And there is no ambush at the end — the auditor is required to report every finding to you before the audit closes, so you leave the closing meeting knowing exactly what was written up. The report itself is drafted on site, then goes back to the certification body for a technical review before it is finalized and made available to you, which the code requires within ten calendar days of the last day of the audit.
Knowing the day runs this way changes how you prepare. You are not preparing to perform for a stranger. You are making sure the records are retrievable, the people can speak to their own work, and the floor matches the binder — because those are the three things the auditor will actually check.
Every finding is a , and the grade decides how much it costs you. SQF Edition 9 uses three grades, and it is worth knowing exactly what separates them:
Those grades feed a score. Under SQF Edition 9 each applicable clause is scored: a minor adds 1 point, a major adds 5, a critical adds 50, and your site rating is 100 minus that total. Land at 70 or above — a rating of "Complies" or better — and, once every finding is closed out, you are certified; below 70 is a fail. There is one shortcut to failing: a single critical non-conformance at an initial certification audit is an automatic failure, and you re-apply. So "passing" is really two conditions at once — a rating that clears the bar, and no finding left open.
Grading is scheme-specific — confirm yours
The 1 / 5 / 50 scoring and the "Complies" threshold above are SQF Edition 9's. Other recognized schemes grade differently — BRCGS, for instance, issues a letter grade rather than a plain pass. Before your audit, confirm how your scheme scores and what rating you need, at the scheme owner's own site, so "passing" means the right thing for the standard you chose. The scheme-choice lesson covers where each scheme's code lives.
SQF does not publish a league table of the most common non-conformances, and you should distrust anyone who quotes you one as a hard statistic. But four failure modes show up again and again for first-timers, and each one maps to a place the code is strict — which is exactly why they are worth pre-empting.
A weak food-safety plan
The Food Safety Plan is a mandatory element of the code, and it is the document an auditor probes hardest. A plan that misses a hazard, names a control it cannot justify, or sets a critical limit with no basis is not a paperwork slip — it is a failure of a system element, which grades as a major. Build the plan so every hazard and every control can be defended out loud, because that is the conversation the audit becomes.
Vague procedures the floor cannot follow
A procedure too loose to carry out the same way twice is, by the code's own definition, an incomplete implementation — a minor waiting to be written up, and one the auditor confirms simply by walking your SOP against the task. The lesson on writing SOPs is the fix: name the what, the how, the how-often, the who, and how each run is verified.
Traceability that cannot be completed on demand
Product Trace is mandatory, and the bar is one step forward to your customer and one step back to your supplier, with the receipt dates recorded. When an auditor asks you to trace a lot and the answer is a scramble through paper — or a chain that dead-ends — that is a system element failing in front of them.
Corrective action that arrives late
Findings are expected; a good audit still produces some. What fails a site is not the finding but the response that never comes, or comes past the clock. That clock is specific enough to deserve its own section.
Getting findings is normal — even a strong audit records a few. What separates a pass from a re-application is what you do with them, and how fast.
A is required for every non-conformance the auditor records. Under SQF Edition 9 the evidence of your corrective actions has to be sent to the auditor, verified, and closed out within 30 calendar days of the last day of your site audit — and for every minor and major finding, that evidence has to include a documented root-cause analysis, not just a photo of the thing fixed. Miss that window, or fail to satisfy the auditor within it, and the certification body cannot certify you; you re-apply. The certification decision itself lands no more than 45 calendar days after the audit.
There is a little give built in. If a finding needs a structural change or a part on a lead time, the certification body can grant more time — but only when the immediate risk is controlled in the meantime and the extension is documented. The lesson is to treat close-out as part of the audit, not a chore that follows it: line up your root-cause process and your evidence trail before the auditor arrives, so a finding on the last day does not become a missed deadline four weeks later.
Everything above is easier if the audit is your second look at these gaps, not your first. The single most useful thing you can do before a certification audit is run one on yourself.
Run a real internal audit against the checklist
The code already requires internal audits, in full and at least annually, against the SQF checklist or a similar tool — so this is not extra work, it is the work the standard expects. Run it as if it counted: record objective evidence, grade what you find, and open a corrective action with a due date for every gap. The SQF Practitioner lesson covers who should run it and why the person auditing a function should not be the person who runs it.
Run a full mock recall
A recall test — reviewed, tested, and verified as effective — is expected at least annually, tracing incoming materials one step back and finished product one step forward, across different shifts. Do it before your audit, on a real lot, and time it. If you cannot complete the trace, you have found a major finding on your own terms instead of the auditor's. The lessons on why you run a mock recall and running one step by step walk the exercise.
Consider a pre-assessment
SQF offers an optional pre-assessment audit — a health check that surfaces gaps before you engage your certification body for the real thing. It is not required, and it costs you either time or a fee depending on who runs it, but for a first-timer it turns unknown risk into a punch list you can work.
A mock recall is only painful when you have to reconstruct the trace — pull paper receiving logs, match lot numbers by hand, and hope the movement history is complete. There is one honest place a tool helps here, worth naming plainly: because Bettr Manager records each lot as it is received, with its supplier lot and expiry, and every movement after, the forward-and-back trace an auditor asks for is already captured, retrieved rather than rebuilt. That is the whole of the mention — the discipline of contemporaneous records is what carries the audit, whatever you keep them in.
Put it together and the first audit stops being a cliff. It is two parts — your records, then your floor — grading findings on a known scale, with a 30-day clock on your response and a self-run version you can do first. A site that has written its plans to be defended, matched its SOPs to the floor, and traced a real lot before the auditor asked walks in to confirm a system that already works, not to find out whether it does.
That is also the bridge to what comes next. Certification is not a one-day performance; it is proof that your operation runs cleanly enough, day after day, to survive a stranger reading your records against your floor. Building that daily operating model — lots, expiry, holds, batch records, and the traceability underneath all of it — is what the running-production track is about.
Do this in your operation