A trading company with a good website and a real factory look identical from the outside. The difference only shows up in documents, in the batch record, and in what happens when you ask for something the seller did not expect. The list below is ordered so that the cheapest checks come first — several of them will end the conversation before you spend money on a sample.
1. Establishment registration, not “approval”
In the United States, a dietary supplement facility registers with FDA as an establishment; the registration is a listing, not a product approval, and no certificate should describe it as one. Ask for the registration number and verify the facility name and address match the entity you are paying. A mismatch between the registered address, the invoice entity and the shipping origin is the single most common sign of a broker in the middle.
2. Who actually owns the equipment
Ask for a photograph of the line running your dosage form with a dated placard, then ask a question only an operator can answer: what is the current changeover time between two flavours, or how many cavities the mould has. Trading desks relay answers slowly and generically; a production manager answers in specifics.
3. Scope of the quality system
A GMP certificate is only meaningful if its scope covers your product type. A certificate covering solid dosage forms does not cover liquids, and a certificate for cosmetics does not cover ingestibles. Read the scope line, not the logo.
4. The last twelve months of COAs
Ask for a COA for a batch you can identify, then for the previous three batches of the same product. Results that are identical to two decimal places across batches are a warning, not a comfort — real analytical variation exists. Check that the specification column, the result column and the method column are all populated, and that the batch number on the COA matches the batch number on the carton.
5. A blank batch record
This is the fastest way to tell a manufacturer from a repackager. A real batch record has weigh tickets, in-process checks with signatures, yield reconciliation and a deviation section. Ask to see one with the commercial details redacted. If the answer is a certificate instead of a record, you have your answer.
6. Laboratory scope
Ask which tests are run in-house and which are sent out, and to which laboratory. In-house capability for identity and assay is normal; heavy metals by ICP-MS and microbiological testing are often outsourced. Neither answer is disqualifying — an evasive answer is.
7. Allergen and cross-contact control
Walk the allergen matrix and the cleaning validation behind it. If the facility runs shellfish, soy, milk or sesame on the same line, ask what the changeover procedure is and whether the allergen matrix is reviewed per batch. A “may contain” statement is not a substitute for a cleaning procedure.
8. Change control
Ask in writing: what happens if a supplier of your active ingredient changes its source, mesh size or carrier? A manufacturer with change control will tell you that you get a notification and, depending on the change, a new stability check. A trader will tell you nothing changes.
9. Traceability from carton to lot
Pick a finished carton and ask them to trace it back to the ingredient lot. This exercise takes ten minutes in a factory with records and cannot be faked in a warehouse without them.
10. Capacity versus your forecast
Ask for monthly output by dosage form, and for the largest single order shipped in the last year. A supplier whose capacity is barely above your launch volume will deprioritise you the moment a bigger customer arrives.
11. Packaging and artwork workflow
Who owns the dieline, who proofs the label, and how many rounds of artwork changes are included? Confirm whether the printer is in-house or external, because that determines whether a label reprint adds three days or three weeks.
12. Two references you can actually call
Ask for two customers who launched a similar product in the last eighteen months. Then ask those customers one question: what happened the first time something went wrong? Delivery performance is easy to be lucky with; problem resolution is not.
What the twelve checks add up to
None of these checks requires a formal audit programme. Six of them can be completed by email in a week, and the batch record and traceability exercises can be done during a sampling visit. Together they answer the only question that matters before a deposit: if something goes wrong in month four, is there a system to catch it, or only a person to apologise?
We run sampling in 3-5 days, ship production in 7 days and issue a COA with every batch, from 100 units. If you want to test these twelve points against us, start with a sample request and a batch record — ask for both at once and see which one arrives first.