Find out what NSF/ANSI 455-4 will say about your plant, before the auditor does.
Send us your procedures, your qualification and validation packages, and a sample of your records. We read the whole standard against all of it and tell you what is wrong, while there is still time to fix it.
Part 211 is a rule about operations. This standard adds a management system on top of it, out of the ICH guidelines. Plants that pass FDA inspection can be missing the whole upper layer.
The standard behind “we need you GMP certified.”
It is what an over-the-counter drug manufacturer gets audited against when a retailer, a customer or a distributor asks for certification. Most people meet it as one line in an email and assume it is Part 211 with a certificate stapled on.
It is not. Clause 1.2 says the standard audits your OTC product to 21 CFR Part 210, to the ICH quality guidelines Q1, Q7 and Q10, and to additional retailer requirements layered on top. That middle piece is the one people miss. Part 211 tells you how to run the operation. ICH Q10 asks whether there is a pharmaceutical quality system sitting above the operation, with management accountable for it.
You can be inspected by FDA, come out clean, and have nothing that answers the ICH half at all — because FDA was never asking.
And it is not only Part 211.
Section 2 names the rules and guidelines that become requirements of this standard by reference. An auditor can write you up under any of them.
The base, cited section by section through the standard — quality unit, personnel, facilities, ventilation, plumbing and the rest.
Clause 4.2.2 wants a quality manual describing the system, the quality policy, management responsibilities and quality risk management. Part 211 asks for none of that.
Named directly in scope, so the expectations they carry are in play whether or not you have ever read them.
If your batch records, training records or approvals live in software, this is in scope. It is the one almost nobody has read.
Clause 4.6.15 pulls the recall machinery in and puts a yearly mock-recall obligation on you with numbers attached.
The standard exists to audit Part 210 and ICH plus what the retailers wanted added. Meeting the federal rule is not the whole test.
Your lab counts too, whoever runs it. A contract laboratory is a purchased service with GMP impact, so it falls inside clause 4.5.2.1 — which wants an approved supplier list, a documented evaluation and selection process covering GMP aspects, a quality agreement defining who is responsible for what, and documented periodic monitoring and review of how that supplier actually performed. Sending work to an accredited lab is how most people evidence that. Almost nobody has the agreement and the performance review.
NSF/ANSI 455-4 System Assessment
We read the whole standard against your system, tell you what is wrong with it, and tell you what to fix first.
And it is built to survive more than one auditor’s reading. This is not one person forming an opinion about your system. Every clause is examined against the documents meant to answer it, the records meant to prove it, the technical discipline behind it, and everything elsewhere in your system that could contradict it. Those reads are made independently and then reconciled into one answer, so forty findings that all come from one missing control arrive as one problem with every affected clause attached.
The certification auditor still decides. They simply have far less left to work out for themselves.
What this covers
- Whether what you have already written is good enough, requirement by requirement
- Which of your documents answers which requirement, and where inside it
- Whether anything you hold answers the ICH half, or only the Part 211 half
- Whether your cleaning validation carries what the standard actually asks of it
- Whether the lab you use has an agreement and a performance review behind it
- What an auditor would write up, and the exact thing that would close it
- Which requirements no document can answer, because they are about the building or about what people do
- What to fix first, so you do not do the same work twice
What you get back
Every requirement of the standard, in the order the standard puts them, marked answered or not. Where one of your documents answers it, we name the document and the page. Where nothing does, we say so and leave it blank.
Then the things that are not simply missing. What is there but does not go far enough. What one of your documents says that another one contradicts. What a procedure claims happens that your records do not show happening.
When forty findings all come from one missing procedure, we write it once and point the rest at it. You get the cause, not the pile.
And we tell you what to fix first. Fix the procedure before you rewrite a year of records to match it, not after. Every entry names the thing that is absent: the evidence, the decision, or the document. Not how to run your plant.
You get the assessment, and a workbook listing every requirement against the document that answers it. Hand that one straight to the auditor.
What you send
Your procedures and controlled documents, your qualification and validation packages, and a sample of your records. We tell you which records and which dates, so you are not guessing and you are not uploading the whole file room.
Five business days, from the day we confirm we have everything. If something important is missing, you hear about it before the clock starts.
What this catches
A plant FDA has never had a problem with, read against this standard. The company is made up. The rules underneath are real and you can look every one of them up.
Every procedure in place, and no quality system above them.
A full set of SOPs, a quality unit with defined responsibilities, batch records reviewed and released. Two clean FDA inspections.
Nothing describes the system as a system. There is no quality manual, no stated quality policy, and nothing that records management committing to any of it or evaluating quality risk.
Clause 4.2.2 wants exactly that document, out of ICH Q10 — the quality management system described, the policy, the organization’s commitment to it, management responsibilities, and quality risk management. Part 211 never asks for it, so a plant built to Part 211 has never had a reason to write it. It is one finding that sits above dozens of others.
Cleaning is validated. The final rinse was never looked at.
Cleaning validation protocols executed on every vessel, swab results within limits, the package approved and filed.
Nothing evaluated whether the chemical residue was actually flushed out of the large vessels at the end — no clarity, pH, odour or TOC on the final rinse water.
Clause 4.5.21.2 wants the cleaning procedures validated and the final rinse water for tanks and large vessels evaluated to show the residue is gone, naming water clarity, pH, odour and TOC. Swabbing a surface and testing what came off the vessel are two different questions, and the standard asks both.
A recall procedure nobody has ever run.
A written recall procedure, a named coordinator, distribution records that could be pulled if anyone asked.
No traceability exercise has ever been run, and nothing sets how often one should be or what result would count as passing.
Clause 4.6.15 wants traceability and mock recall exercises at least once a year, forward and backward, landing between 99.5% and 101.5% recovery inside four hours, with any additional exercises finished inside twenty-four. Those are numbers, and either your last exercise met them or there was no last exercise. On paper, a procedure nobody has run looks exactly like one that works.
Made-up company, real rules. We write these to show you the kind of thing that gets found, not to claim we found it. Your assessment is written against your own documents.
Price
What actually lands on your desk.
Four pieces of a finished assessment, opened here on the page. They come out of one written against a company we invented and read against NSF/ANSI 455-2, so you can see the shape of the thing before you pay for it. The company, its documents and its records do not exist. The requirements underneath are the real, published requirements of that standard.
Cascade Nutraceuticals is invented, and so are its documents and its records. They were written to show what the deliverable looks like, not to claim a catch. This specimen is an assessment against NSF/ANSI 455-2, not against NSF/ANSI 455-4: what changes from one scheme to the next is the standard behind the read, not the shape of what comes back. The requirements it is read against are the real, published requirements of NSF/ANSI 455-2. The standard’s own wording is not reproduced here; the clause identifiers and the federal citations are.
Every requirement of NSF/ANSI 455-2-2024, in the standard’s own clause order, with how it is answered, what that rests on and where inside it, and any finding against it. Five of the 695 rows.
The delivered workbook carries each requirement in the standard’s own wording. That wording is not ours to publish, so what you read below is our paraphrase of it. The clause identifiers and the federal citations are the standard’s own.
| Clause | The requirement, paraphrased | How it is answered | What that rests on, and where inside it | Finding |
|---|---|---|---|---|
| 4.1.1 | Whoever holds quality responsibility is not the post that runs operations. [21 CFR 111.12(b)] | Not found in what was supplied | Read against OOS-24-007 signature line, HR-TRN-01 “Competency by Function” first row, and LAB-2403-118 “Review & Approval”. Each of the three records the quality post and the operations or analyst post as one post; none establishes the separation the requirement calls for, and no organization chart, written assignment or delegation record was supplied. | 4.1.1 |
| 4.1.2 | The quality control work is identified, and it is carried out. [21 CFR 111.65] | Addressed, not settled from what was supplied | SOP-QC-012 §3.0 identifies the quality control duties, and OOS-24-007 and LAB-2403-118 record those duties being exercised on two lots. What was supplied does not reach whether quality control operations are implemented across manufacturing, packaging, labeling and holding: no batch production record, master manufacturing record or packaging and labeling record arrived, and none was declared as not existing. | No finding |
| 4.2.1 | Written procedures set out what quality control is responsible for. [21 CFR 111.87, 111.103, 111.105, 111.140(b)(1)] | Evidenced, no finding | SOP-QC-012 §3.0 Responsibilities. | No finding |
| 4.4.19 | Sewage and waste are plumbed away from the plant so they cannot reach contact surfaces, product, components or the water supply. [21 CFR 111.15(g)] | No document can answer it | Named at the front, under what documents could not answer. It turns on how the building is plumbed, which nothing you could send can settle, and no directed remote observation was purchased. It carries no finding. | No finding |
| 4.6.15 | Quality sets and approves the laboratory controls, including the sampling plans, how specifications are set, the test methods and the reference standards, and those methods are used to the criteria set for them. [21 CFR 111.315] | Evidenced, with a finding | LAB-2403-118, “Reference Standards & Reagents”; “System Suitability”. | 4.6.15 |
Eight dispositions run down that third column and every one of the 695 rows carries one. “Not found in what was supplied” says nothing was found in what was examined. It never says the control does not exist in your operation.
Constructed specimen. Cascade Nutraceuticals, its documents and its records do not exist and were written to show the deliverable. The requirements are the real, published requirements of NSF/ANSI 455-2. Nothing here is a finding against any company.
Every document supplied and what each one points at, built from the same references the drawn map in the assessment is built from. Fourteen documents arrived.
| Your document | What it references | Was the reference supplied | What points at it | How many point at it |
|---|---|---|---|---|
| SOP-QC-012 | FORM-QC-12A; FORM-QC-12B; SOP-QA-005; SOP-QC-002 | None supplied | OOS-24-007; HR-TRN-01 | 2 |
| MV-RHO/ASH-01 | ASSAY-RHO/ASH-01 Rev. 0; SOP-QC-002 | None supplied | Nothing points at it, and nothing has to | 0 |
| LAB-WS-007 | ASSAY-RHO/ASH-01 Rev. 0; QC-CALC-Rhodiola.xlsx; LAB-2403-118 | LAB-2403-118 supplied; ASSAY-RHO/ASH-01 Rev. 0; QC-CALC-Rhodiola.xlsx not supplied | Nothing points at it, and something has to | 0 |
| LAB-2403-118 | ASSAY-RHO/ASH-01 Rev. 0; QC-CALC-Rhodiola.xlsx | None supplied | LAB-WS-007 | 1 |
| OOS-24-007 | SOP-QC-012; CA-24-006 | SOP-QC-012 supplied; CA-24-006 not supplied | Nothing points at it, and nothing has to | 0 |
| QC-RSV-01 | Record-retention SOP | None supplied | CMP-LOG-AS | 1 |
| HR-TRN-01 | SOP-QC-012; Record-retention SOP | SOP-QC-012 supplied; Record-retention SOP not supplied | Nothing points at it, and something has to | 0 |
| SQF-RHO-03 | Supplier certificates of analysis; RM-RHO | None supplied | Nothing points at it, and nothing has to | 0 |
| REG-CLM-AS500 | AS-LBL Rev. 3 (2023); Ingredient supplier monographs | None supplied | Nothing points at it, and something has to | 0 |
| CMP-LOG-AS | QC-RSV-01 | Supplied | Nothing points at it, and something has to | 0 |
| PDC-AS700-01 | MBR-AS700; MF-AS700 rev A; SOP-VAL-02; CC-25-031 | None supplied | Nothing points at it, and nothing has to | 0 |
| CWC-SQCA-220 | CWC-SCA | Supplied | CWC-SCA | 1 |
| CWC-SCA | CWC-SQCA-220 | Supplied | CWC-SQCA-220 | 1 |
| NNRG-SQ-014 | No document named | No reference to check | Nothing points at it, and nothing has to | 0 |
Sixteen of the twenty-one documents this set names were not themselves supplied. They get their own sheet in the workbook, and that is the sheet worked first: it is the shortest route to closing the largest number of open rows on the requirement map.
Constructed specimen. Cascade Nutraceuticals, its documents and its records do not exist and were written to show the deliverable. The requirements are the real, published requirements of NSF/ANSI 455-2. Nothing here is a finding against any company.
Each finding is headed by the requirement it rests on and a short name for what is at issue. Findings sit in the standard’s own clause order, under the group stating what the assessment established about your record, and each one closes on the consequence that follows from it. This is one of eleven.
The delivered assessment heads the finding with the requirement in the standard’s own wording. That wording is not ours to publish, so the requirement line below is our paraphrase of it.
The release result reported for lot AS4830922 was not calculated against a reference standard that was within the expiry the record itself states for it.
LAB-2403-118, “Reference Standards & Reagents” table: the Rosavin reference standard, lot R04219, carries an expiry of 30 NOV 2023. The same report records the run as acquired 11 MAR 2024 and the reportable result as calculated from the bracketing Rosavin standard mean.
LAB-WS-007 for the same lot repeats the standard lot as R04219 and the reference-standard expiry as 11/30/23, and records the reportable value as transcribed from a spreadsheet named in the worksheet rather than calculated on the worksheet.
LAB-2403-118, “System Suitability”: the replicate-standard %RSD is entered as “not calculated” and the suitability acceptance criterion is entered as “per method file”, so the record carries no criterion the standard was judged against on the day of the run.
The reference standard is what the result is measured against; where its expiry has passed, the number the lot was released on has nothing established behind it.
What was supplied does not establish a supportable result for the release of lot AS4830922.
Constructed specimen. Cascade Nutraceuticals, its documents and its records do not exist and were written to show the deliverable. The requirements are the real, published requirements of NSF/ANSI 455-2. Nothing here is a finding against any company.
One row per finding, in the standard’s own clause order. Eleven rows. This is one of them.
The delivered register carries the requirement in the standard’s own wording. That wording is not ours to publish, so the requirement cell below, and the one place another cell restates a clause, are our paraphrase.
QC-RSV-01, “Reserve Sample Records”, the row for lot AS4815602 (manufactured 11/2023, expiring 11/2026): the units retained column reads “none on file”, and the container, retain-until, storage and logged-by columns are each empty.
QC-RSV-01 is the only record supplied that addresses reserve sample collection; its header states the retention policy as written rather than referring to a procedure, and names no procedure that governs it.
Everything under “Yours to fill” is yours, and it is delivered empty. No owner is assigned, no date is set and no status is pre-filled, because each of those needs to know what your operation can absorb and nothing in this assessment knows that. There is no formula in the file, so you can add columns, add rows and set your own filters without breaking it.
Constructed specimen. Cascade Nutraceuticals, its documents and its records do not exist and were written to show the deliverable. The requirements are the real, published requirements of NSF/ANSI 455-2. Nothing here is a finding against any company.
Straight answers.
You never see my plant. What is the point?
We read every requirement. The ones that are about the building, or about what people physically do, get named as exactly that and get no finding. The rest can be answered from what you send, and that is most of the standard. The walkthrough exists if you want the remainder answered too.
Will you tell me what grade I am going to get?
No, and nobody honestly can. The auditor sets the severity of every finding on the day, and the certification body reviews the report afterwards. What we can tell you first is which requirements your system does not answer.
We are inspected by FDA already. Is that not the same thing?
No. FDA inspects you against Part 211. This standard adds the ICH quality guidelines and the retailer requirements on top, and audits all of it together. Coming out of an FDA inspection clean and still failing this audit is the normal case, not the unusual one.
What if you find something we cannot fix in time?
Then you know before the auditor does, and you get to decide what to do about it. That is the point of the name. We lay out what is there and what would close it; which risks you carry into the audit is yours to choose.
Our audit is not booked yet. Is this useful now?
That is the better time for it. Getting on a certification body’s schedule takes months, and once a date is set they charge to move it. Everything this finds is work you would rather be doing before there is a date at all.
Is what I send confidential?
Yes. What you send is used only to do the work you paid for, seen only by the people doing it, never sold, and never used to train any public AI model. It is kept only as long as we need it. Section 4 of the privacy policy, linked at the foot of every page.
This page is about one standard. You may need something else.
An assessment reads your whole system and tells you what is wrong with it. If that is not the job in front of you, here is where the rest of the work goes.
Regulatory work product, not legal advice. An assessment is prepared from the records you submit and is not legal counsel, not a certification decision, and not a prediction of any audit outcome or grade — those are decided by the certification body. It is a read of documents: no site visit is performed, and every requirement that depends on physical conditions or observed practice is named as outside the read and gets no finding unless the optional facility walkthrough is purchased, in which case each such finding states what was shown, by whom, and on what date. It relies on what you supplied and on your own declarations about your operation, none of it independently verified. Clause references are quoted from NSF/ANSI 455-4. All companies and products shown in examples are constructed; any resemblance to a real company or product is coincidental. Regulatory Options is not affiliated with or endorsed by NSF.