Glass Bottle Quality Certificate: How to Verify It

One of our sourcing guides for glass packaging — each takes one question a buyer has to settle before ordering.

A glass bottle quality certificate proves something about your goods only when it is tied to a batch: it names a lot number that is also printed on the delivered cartons, the quantity it covers, the production date, the sampling basis, and measured values set against limits and methods. To verify it, start from the goods instead of the paper. Pick a carton at random, ask for the certificate and the retained sample for that exact lot code, then re-measure one reported value with the plant's own gauge. A document that fails those checks is a general statement about a product family, however official it looks.

Three document types that get called a certificate

"Quality certificate" is a commercial phrase, not a defined document. Suppliers use it for three different things, and most paperwork disputes begin when a buyer asks for one and receives another. Before judging the content, establish which of these you are holding.

For glass packaging, the characteristics on a certificate of analysis that normally carry numerical limits are capacity, overall height, body diameter, weight, neck finish dimensions, internal pressure resistance, thermal shock resistance and, where the market calls for it, food contact migration. A certificate showing only a visual pass says little. Visual acceptance is the weakest kind of evidence and the hardest to reproduce from one inspector to the next.

Test reports come in three kinds that can look the same on paper. A type test, run once on a prototype, shows what the design can do. A periodic test, on samples pulled from routine production, shows how the process was behaving at that moment. A batch test, on samples drawn from one lot, is the only kind that can stand behind a certificate of analysis for that lot.

A declaration may rest on test reports the signatory holds but does not attach, or on a documented assessment. Because it usually binds the person who signs it, nobody should sign one casually, and a buyer should not accept one in place of results. The reverse also applies: in several markets a pile of test reports with no declaration is as useless as a declaration with nothing behind it. How the declaration and the wider compliance package are built, market by market, is set out in our guide to the food contact declaration for glass bottles.

Who is asking decides which one you need

• The brand's quality department has to release the goods, so it normally wants a batch-linked certificate of analysis with measured values.
• A retailer's technical file normally wants a declaration together with the test reports that support it.
• The filling plant wants the dimensional results in a form it can transfer to a line setting sheet.
• A customs broker wants whatever the destination's import rules prescribe, in the right language and with the right authentication. That is usually a question of format, not of technical content.

Settle which of these four the request comes from before the order is placed. Otherwise a certificate prepared at loading can be turned away on arrival for being the wrong type, even though nothing in it is incorrect.

What a usable certificate of analysis contains

A certificate of analysis that holds up under scrutiny has eleven elements, plus the raw data that sits behind them. Each one closes a specific gap, so a document lacking two or three of them is unusable, not merely thin. The table lists each element with the form a credible version takes, the way it typically fails or is fabricated, and a check an auditor can run on site with no laboratory.

Elements that need a closer look

Revision. A certificate naming the product without the revision can be technically valid and still mislead, because the delivered bottles may follow a later revision with different limits.

Sampling basis. This is the element most often absent altogether. A result taken from five pieces and a result from a properly run ISO 2859-1 plan at a stated AQL are indistinguishable once the sampling section is left off, yet only the second says anything about the lot.

Method. An in-house method is acceptable provided it is written down, carries its own number and revision, and can be repeated. A method that lives only in verbal practice cannot be audited, and its figures cannot be set beside anyone else's.

Dates and signature. The date of issue is not the date of test, and both belong on the page. An unsigned certificate, or one carrying the scanned signature of a former employee, is worse than a blank form because it creates a false sense of protection.

Accreditation. A laboratory accredited to ISO/IEC 17025 carries more weight, but accreditation is granted test by test. The question worth asking is whether the particular test on the report sits inside the accredited scope.

Conclusion. The certificate should state acceptance or rejection against the limits it lists, and say what happens to a lot that is rejected.

Where a market sets its own requirements, such as food contact testing under EU 10/2011 or compliance statements under FDA 21 CFR, those results should trace to the samples that were certified. A test run with a different simulant or under different conditions has to be labelled as such and not offered as equivalent.

Checking the certificate against the physical lot during an audit

The dependable way to learn whether a certificate programme is real is to move from goods to paper, not from paper to goods. It takes about an hour, needs no laboratory, and cannot be rehearsed as long as the auditor chooses the sample.

1. Start in the finished goods warehouse, not the meeting room. Walk the storage area and pull two cartons yourself, from different positions in different stacks, without asking which lots have certificates. Write down each lot code exactly as printed, with any line or machine identifier and the production date if shown. A file assembled for the audit holds the lots someone else selected; a random draw is not among them.
2. Ask for the certificates covering those two codes. Let the plant fetch them from its normal filing system and watch where they come from. A document pulled promptly from a maintained electronic index points to a working system. One that turns up after a wait, on freshly printed paper, was produced by the audit and not by production.
3. Ask for the retained sample for each code. Retention samples are the physical twin of the certificate. Check that the sample bears the same code, is sealed and labelled like the cartons, and sits in a store arranged by date and not in an unsorted heap. If the plant keeps no retained samples, ask how a complaint about one delivered batch would be investigated, and note whether you hear a procedure or an improvisation.
4. Measure one bottle. Take a bottle from the chosen carton and, with the plant's own gauge, measure a characteristic the certificate reports as a number. For glass that is usually capacity, overall height or a neck finish dimension. If the reading lands clearly outside the certificate tolerance, the document is decorative. No other check is as decisive, since it uses the plant's equipment and leaves no room to blame the auditor's instruments.
5. Trace the lot code into the production records. The code should decode to a real production date, a machine or line, and a quantity that appears in the production log.

Two anomalies deserve a note in the audit report. One is a run of consecutive lot numbers with gaps that match no stoppage, which suggests numbers are handed out for convenience. The other is handwritten entries on a certificate or test sheet with no initials and no instrument output beside them.

Ask for the revision history of the file as well. Certificates for older lots that cite a superseded item code revision are normal when the drawing changed. They become a problem when no new certificate followed the change.

If the market requires food contact documentation, run the same goods-to-paper check on the compliance file and ask whether the test results trace to the specific lot or only to the product family. All of this depends on a lot code that can be followed through production, which is the subject of our page on batch traceability for glass bottles. Inspecting the bottles themselves before loading is a separate job from reading a document about them; that work is described under pre-shipment inspection of glass bottles.

Why certificates from different suppliers cannot be ranked side by side

Certificates from three candidate plants are written to three conventions. Reading them line by line gives a false ranking unless you first check the six points on which they differ.

• Test scope. One plant certifies dimensions and capacity because its customers have always asked for those. Another certifies heavy metal release for its main market. A third certifies internal pressure and thermal shock because its bottles go through a pressurised process. Each calls the result a quality certificate, and each has tested what its own customers demanded, not more or less thoroughly than the others.
• Sample basis. Type, periodic and batch tests can share a format and a year, and only the batch test speaks for the goods. A supplier who has to look up which kind a report was has a document control problem, not a testing problem.
• Sampling plan. Identical measured values can rest on five pieces or on a properly sized sample at a defined acceptance level. The confidence behind them is entirely different, and a certificate that leaves the plan out cannot be weighed against one that states it.
• Attributes versus results. "Accepted" says nothing about how near the lot ran to its limit. Measured capacity, measured finish diameter and the pressure at which each sample failed can be trended and matched against the buyer's own readings. Two suppliers reporting acceptance look equal; two reporting figures can be ranked.
• Method and conditions. The same bottle gives different pressure figures at different rates of load application. Migration into water, dilute acetic acid and a fatty food simulant are three exposures with three results on the same glass, and for some analytes one is far harsher. A drop test with no stated height is not a test. A number with no method cannot be compared with anything.
• Age and revision. A report from an earlier year may describe a different mould, glass supplier and colourant batch. Older data still has value as background if it is labelled historical, but the current certificate must cite the current revision.

Where these differences appear, stop trying to rank the documents. Send every candidate one document specification that fixes the document type, the characteristics, the limits, the sampling basis, the language and the signatory. The answers become comparable once every supplier has been asked the same question. When we match a project to a plant, this specification goes out with the drawing so that the paperwork is agreed before samples are made.

Responding to customer, customs and tender requests for documents

A request for paperwork is one of three kinds. A customer's technical request concerns evidence and usually names a market and a product category. A customs or import request concerns format and authentication under the destination's rules. A tender or retailer onboarding request concerns completeness and is met with a pack of documents, not a single one. Answering before you know which kind it is tends to produce an excellent document in the wrong form, rejected twice.

Technical requests

Reply first with a question, not an attachment: which market, which product category, and which document type the requester's own procedure names. A request worded as "certificate" may mean "declaration" inside the requester's system, and the two are not interchangeable. With the type fixed, the market's requirements set the characteristics to cover and the requester's quality procedure sets the sampling basis.

Customs and import requests

Here format dominates. Depending on the destination, the document may need company letterhead, an original signature, a chamber of commerce stamp, translation into the destination's official language, or legalisation by a consulate or apostille. The importer of record or the destination broker decides which applies, not the supplier. Ask for the exact requirement to be confirmed before production, so the document is prepared together with the goods. Translate the entire document, method references and limits included; a translated header above an untranslated body is a frequent cause of a second rejection.

Four statements to keep apart

• A batch certificate of analysis can be issued for delivered lots.
• A declaration of conformity can be issued from the composition and the test evidence held.
• A third-party test report can be obtained for a specified test on specified samples.
• The manufacturer holds a product certification or approval mark.

Only some of these hold for any one plant. Claiming a certification that is not held is the most damaging paperwork error there is, because the requester can check it with one message. When a document cannot be supplied, name the nearest alternative and give the reason instead of passing over the requirement in silence.

Keeping the certificate file valid as the product changes

Documents that were accurate on the day of issue weaken as the product and process shift around them. The decay usually goes unnoticed until a customer asks something the file cannot answer.

The changes that undermine existing documents are predictable, and the supply agreement should list them as notification triggers:

• a different glass supplier or furnace
• a different colourant, or a switch from one amber to another
• a new or repaired mould
• a change of capacity or neck finish
• an added decoration process such as organic coating or ceramic printing
• a change to the closure or liner
• a change in the destination market's regulation

Any of these can alter a result recorded by an earlier test, and none shows up for a buyer who sees only goods and certificates. A change notification clause, with the right to re-qualify after a notified change, turns the file from a snapshot into a maintained system.

Two habits keep the file usable. The first is a shared document index: a single list, held by both sides, that links item code, drawing revision, document number, revision, issue date and the lot range each document covers. Without it, a buyer cannot tell whether the certificate in hand is the current one and will tend to assume that it is. The second is an expiry rule for type and periodic tests written into the agreement, so that a date or an event prompts the re-test and not a dispute.

Keep certificates and their raw data for at least as long as the product may still be on the market. For containers that period is longer than for many consumer goods, since bottles are often filled and shipped well after delivery. Electronic issue is better than paper provided the file cannot be edited, every document carries a number the issuer can confirm, and the route for confirming it survives staff changes. When a signatory's departure makes a certificate impossible to verify, the remedy is a document control procedure, not a fresh signature. The containers these files relate to are listed in our glass bottle and jar collections.

Questions buyers ask about glass bottle quality certificates

Is a quality certificate the same thing as a certificate of analysis?

Not always. Suppliers apply the phrase to batch certificates, third-party test reports and declarations of conformity alike. Ask which of the three is on offer, what document number it carries and whether it relates to the delivered lot. Only then can you tell whether it supports a claim about your goods.

How do I know a certificate was written for my batch and not for a generic product?

Check for three things together: a lot number that matches the cartons, the quantity covered and the production date. If any is missing, the document describes the product family. If all are present, test the link by taking a carton at random and asking for the certificate and retained sample for that code.

Can I rely on results from an accredited laboratory?

Accreditation helps, within limits. ISO/IEC 17025 accreditation covers a defined list of tests, so a laboratory may be accredited for one and not for another. Confirm that the test on your report is inside the scope, and whether the laboratory performed the work or subcontracted it. A mark shown without a scope statement tells you less than it seems to.

Is a test report that is several years old still worth anything?

It stays useful as background on the design and loses force as a statement about current production, since the glass supplier, colourant batch, mould or drawing may have changed. A type test should be repeated when a defined change takes place, and a periodic test refreshed at a stated interval. Ask which kind the report is and which events would invalidate it.

My customs broker wants a "certified" document. What is being asked for?

In most cases a format, not a technical content: letterhead, an original signature, a chamber of commerce stamp, a translation or consular legalisation. The importing country defines these, so the reliable answer comes from the importer of record or the broker. Get it before production begins.

What should I send to get a useful answer about documents?

Two items matter most: the destination market, which determines the regulation, and the list of tests your buyer or quality procedure requires. Add the product category, the document type your process names, and any requirement for a particular language, signature or authentication. From that we can say which documents are relevant, what each should contain, what sampling basis a batch certificate should rest on, what to ask when the same requirement goes to more than one supplier, and how to check the documents against delivered lots.