A pre-shipment inspection of glass bottles is run in the gap between a finished, packed batch and a closed container, and it only works if five things were fixed in writing before production began: when the visit happens, how the sample is drawn, how each fault is classed, what the report must contain, and what follows a failure. Sampling mechanics come from ISO 2859-1 or ANSI/ASQ Z1.4, but the acceptance limits are set per order and per defect class, never copied from another job. Tie the balance payment to a compliant report on a frozen lot and the inspection becomes a release decision instead of a visit.
Two related questions sit outside this article. Picking between inspection routes and working with an agency is covered under third-party inspection for glass bottles. Judging whether a test report or certificate is authentic, and what it ought to say, belongs to our notes on the glass bottle quality certificate. Here the route is taken as chosen, and every sampling value mentioned is a reference approach to be agreed against the order's risk: the contents, the destination market, the filling line and the damage one bad bottle could do in a consumer's hands.
What the inspection decides and what it cannot fix
Think of the inspection as a gate. A supplier's assurance about a batch is replaced by a defined lot with a recorded result, and a commercial decision is then taken on that record. Buyers who let it slide into a formality tend to learn the true state of the goods on their own filling line, after their bargaining position has gone.
By the time the inspector walks out, three things should be known:
- whether the finished lot meets the specification, judged on written characteristics and tolerances and not on a general impression;
- whether quantity, packing and marking match the order, down to carton count, pallet pattern and label content;
- whether the recorded result is solid enough to release the balance or to hold it back.
An inspection that leaves the last point open has gathered data but delivered no decision.
Its limits are just as firm. It does not stand in for in-process control: checking a batch nobody watched through forming and annealing simply reveals a big loss at the stage where every remedy is costly. Nor is it a design review. A bottle drawn wrongly for the filling line will pass as a consistent copy of the wrong drawing, because inspection confirms conformance to a specification and has no power to write or repair one.
The five items to fix in writing before production
An inspection plan has few moving parts. Settle each on paper ahead of the first mould run and the inspector's findings bind both parties; leave them open and each finding turns into a debate on the day.
- Timing. The point in the order at which the visit takes place.
- Sampling plan. How the lot is defined, which inspection level applies, and the acceptance quality limit for each defect class.
- Defect classification. The rules that decide how a given fault is counted.
- Report. Its mandatory content and its turnaround.
- Disposition. What a failed lot leads to, who can order sorting, who can grant a concession, and who pays in each case.
Name all five and you have a plan. Name only a date and you have a booking.
The protocol that carries these items is written and countersigned before the first mould runs. It lists the characteristics to check, their tolerances, the sampling level, the limits per class, the instruments, the viewing conditions, the report format and the disposition rules. If you already hold a written quality standard, attach it to the purchase order; issuing it once the goods exist is too late.
Timing the visit: production completion or loading
The window opens when production and packing are both finished and closes when loading starts. Where inside that window you place the visit is a commercial choice more than a scheduling one, because it decides which remedies remain.
| Inspection point | What it examines | Remedies still open after a fail | Condition or weakness |
|---|---|---|---|
| Production completion | The finished lot while the supplier still controls it as a lot | Sorting, rework, re-running a mould or a negotiated concession, all without disturbing the shipping schedule | The inspected goods must be set aside, identified and frozen, with nothing added later |
| Container loading | Exactly the goods entering the container, with their packing, carton count, pallet pattern and marking | In practice only rejection, and the schedule suffers first | The result comes at the last moment, so both the sample and the remedy are constrained |
For large or higher-risk orders, split the work in two. Do the substantive inspection once production is complete, then add a shorter check at loading that confirms quantity, packing and lot identity against the first report. When the risk lies in the packing more than in the glass, our guide to packing and consolidating bulk glass containers goes further into the loading side.
Freezing is what holds the completion-stage inspection together. If the plant goes on producing or sorting after the inspector has left, the report describes a sample that no longer matches the shipment and stops being evidence of anything. So freeze the lot in writing when the visit opens, log the pallet and carton references that were inspected, and require any later addition to be declared.
Building the sampling plan around order risk
ISO 2859-1 and ANSI/ASQ Z1.4 give both sides a shared vocabulary: inspection level, AQL, switching rules. Use it. What neither scheme does is pick your risk level. A decorative jar for a domestic gift channel and a pharmaceutical container should not carry the same acceptance quality limit, and a supplier quoting one fixed figure for every order is offering convenience, not control. The test is what happens downstream when a defective unit slips through. A leak, a broken filling head, a rejected retail carton and a safety incident are four separate consequences, and each justifies its own degree of tightness.
Inspection level
Normal practice works from a general level. The middle one is the default, and the tighter ones are kept for orders where a single bad unit is expensive later on. Stepping up enlarges the sample, which raises confidence and also raises inspection cost and time, so let the consequence of accepting a bad lot drive the choice, not how hard either side pushes. Safety-related characteristics can be handled apart from the main inspection under the scheme, and that is the proper treatment for any fault capable of injuring someone.
A separate limit for each defect class
Write three limits, not one. Of everything a buyer can do at this stage, this pays back most.
- Critical: effectively zero tolerance in nearly every commercial setting, since a critical defect can cause harm or make the product illegal to sell.
- Major: governed by what the filling line and the consumer will put up with. It gets tighter as market expectations rise and as a downstream failure becomes more costly.
- Minor: can be relatively generous, because these faults are cosmetic and show only under close examination.
Lot definition and stratified drawing
Sampling applies to a lot, so the lot needs a definition first. The expensive mistake is to call an entire order one lot when it spans several production days, moulds or colour batches. A fault confined to one mould gets diluted across the full quantity; it may escape the sample altogether, or show up at a rate that understates how bad it is.
Define the lot as one production run of one model, one colour and one decoration. Then spread the sample over production times, moulds and pallet positions, and have the drawing method written into the report. A result obtained this way points to a cause, not just to a count.
Destructive checks
Separate the checks that leave a bottle intact from those that may not. Dimensional and visual checks are non-destructive; a capacity or strength check can consume the unit. Destroying a small number of bottles, as in a drop test, is a legitimate input to a release decision, yet it is not attribute sampling and needs its own line in the protocol with method and sample size stated. Our page on the drop test for glass bottles describes the general approach where impact resistance forms part of acceptance.
Grading glass defects by consequence
Glass has a familiar family of faults. Grade each by what it does downstream, not by how easily it is spotted on the factory floor. The grouping below is a starting position for negotiation; boundary values such as a permitted bubble diameter or a maximum chip length are agreed for the specific product and market.
| Fault | Usual class | Reasoning | What to agree |
|---|---|---|---|
| Breakage, through-wall cracks | Critical | A cracked wall or neck gives way under filling or transport stress and can injure whoever handles it | That more than an incidental rate triggers a process investigation, not a sort |
| Chipped mouth, damaged sealing surface | Major | The closure no longer has an unbroken surface to seal on, which leads to leakage, cap-slip or a failed integrity test | The finish as a functional characteristic for crown, lug and threaded closures |
| Stones (unmelted or partly melted batch material) | Major in the body wall; critical on or near the sealing surface or when large | A stone weakens the wall and can act as a stress raiser out of proportion to its visible size | Grading by position and size |
| Bubbles, blisters, enclosed gas voids | Minor to major | A lone small seed in a thick base matters little; a cluster of larger voids near the shoulder sits where axial load is carried during capping | A diameter threshold, a count threshold per unit and an exclusion zone |
| Lean and other dimensional departures | Major when the downstream process is affected | The bottle jams a labelling station, mis-seats on a filling carousel or packs badly | A tolerance and a measuring method for each characteristic |
| Light scratches, in-tolerance mould seams, oil marks, dust, slight colour variation within range | Minor | Visible on shelf but seldom stops a filling line | The agreed colour range and seam tolerance |
Cracks deserve a further note. Those that begin at the finish or the shoulder are the worst kind, since they cross the sealing surface and the most highly stressed zone of the container. When cracked units appear above an incidental rate, suspect forming or annealing before you blame handling.
Lean is a good example of why measurement before shipment matters: unless verticality was checked at the plant, the jam it causes later will be blamed on the filling line. Capacity, bore and thread dimensions, wall thickness distribution and base flatness follow the same reasoning. Give each a stated tolerance and an agreed method, because a nominal value with no tolerance cannot be inspected at all.
The trap throughout is grading by visibility. A chipped finish looks slight and is functionally serious. A small scuff on the base catches the eye and harms nothing.
How the visit runs on site
Most inspection failures are failures of order, not of effort. While the batch is being made, the supplier's end-of-line checks should run against the same protocol the inspector will use, so that the later visit confirms a controlled process instead of attempting a rescue. Readings and batch references from those checks join the evidence the inspector reviews.
On the day, the sequence is:
- Identify the lot and confirm that no production or sorting is under way.
- Draw the stratified sample, not whatever stands nearest the door.
- Measure dimensions with the agreed instruments and check capacity by the agreed method.
- Assess visual faults under the agreed lighting, not the light a loading bay happens to provide.
- Check sealing-related features wherever the closure falls within scope.
- Photograph faults against a reference, so a disputed unit can be discussed later without shipping it back.
The report follows within the agreed turnaround. The lot is then formally released, held or rejected, and the payment trigger in the contract fires or stays put. That written close-out is the difference between an inspection and an opinion.
What a usable report contains and how to read it
A report earns its keep when you can argue from it. At a minimum it should show:
- lot identification, with quantities presented and inspected;
- the sampling scheme that was really applied;
- each characteristic's result, including the underlying counts;
- defects tallied by class;
- instruments used and their calibration status;
- lighting and measurement conditions;
- photographs of the faults found;
- inspector's name, date and time;
- the decision and the basis for it.
A bare "pass", or a percentage with no statement of what was measured, supports neither a claim nor a payment release.
Read it against the protocol, not against your hopes. For every characteristic, ask one thing: does the result fall within the agreed limit? A value that passes but sits near the edge tells you something about the process and is a topic for the supplier conversation, not grounds for rejection. A value that fails where the class looks misapplied is settled by comparing the unit with the reference photographs on the agreed list, which is the whole reason to negotiate illustrated boundary cases, one acceptable and one rejected example of each contested fault, before production.
Keep the document types apart as well. An inspection report is not a test report and not a certificate, though the three are often mixed up.
Options when the lot fails
Three routes open after a failed result. Each has a cost, and each is easier to live with if it was priced into the protocol beforehand.
| Route | When it fits | Trade-off accepted | Discipline required |
|---|---|---|---|
| Rework or full sorting | Keeping the shipment and the schedule matters and sorting is cheap next to the downstream cost | Labour and time, handling damage during the sort, and the danger that failed units are quietly swapped for bottles from an unexamined source | Re-inspect the sorted lot before release; without that, the sort is only a promise |
| Concession (use-as-is) | The defect is real and over the limit, but its downstream effect is tolerable for this order and end use | The downstream risk itself: slower filling, a visible share of imperfect units at retail, or a later complaint | A written deviation naming the affected characteristics and the quantity |
| Rejection | The standard has to be protected even at the expense of the schedule | The schedule | Use it while the balance is unpaid and the goods have not left the plant |
Rejection is the strongest remedy, and it has a short life. After the container sails it turns into a claim, and the buyer's position is far weaker.
Choose by weighing the sorting cost against the cost of living with the fault. Whatever the answer, it should not be improvised on the quay under time pressure. Pre-agreeing the options, the evidence each one needs and the party who pays makes the decision procedural.
Linking each stage to the balance payment
The plan should connect every stage to a commercial hook. Without that link, a finding is merely an observation.
| Stage | Typical dispute | Payment condition to write in |
|---|---|---|
| Timing | The tail of the order is finished after the visit and shipped unexamined | Release depends on a report that covers the frozen lot reference, not on the existence of some report |
| Sampling | Several production days are lumped into one lot, or the sample comes from a single pallet | Sampling disputes are settled against the signed protocol before any payment |
| Defect grading | The buyer says major, the supplier says cosmetic, usually because the defect list had no pictures | Terms refer to the class totals in the report, so reclassification is argued before payment |
| Report | It arrives once the vessel has sailed, or gives results without counts | The trigger is receipt of a compliant report, not completion of the visit, and the loading gate stays shut until it is issued |
| Disposition | Sorting is agreed informally, then contested when the labour invoice or the delay appears | The balance is held until the chosen route is closed out with evidence |
The rows form one chain: define the lot, sample it fairly, judge against an illustrated list, record the outcome, settle who pays for a failure. Arguments usually arise because the chain was assembled in reverse, with the visit booked first and the protocol drafted afterwards.
Mistakes that undermine the result
Disappointing inspections fail in a handful of repeatable ways, all preventable at the planning stage:
- a sample taken from one pallet or from the top layer of a stack;
- production or sorting that carries on once the inspector has gone, so unreported goods end up in the shipment;
- a specification giving nominal dimensions without tolerances, which makes any measurement impossible to judge;
- defect classes still undecided on site, stalling the visit and yielding a report that hedges;
- a missing or uncalibrated instrument, which strips out the numeric part and leaves visual opinion;
- an unmarked, unfrozen lot, so that nobody can later say which goods were examined.
These are commercial failures more than technical ones. They appear when the inspection is organised as a visit and vanish when it is organised as a control point.
What we need to draft an inspection plan
Three inputs are enough for us to draft a plan with you: the bottle type or drawing, the batch quantity with its production schedule, and whatever requirement your own customer has placed on the order. From these follow the lot definition, a suitable inspection level, the characteristics worth measuring, the defect classes that matter for the end use and the payment-linked deadline for the report. An existing specification, quality standard or customer checklist speeds things up, since acceptance criteria can be written straight against it. Figures in the resulting plan remain a reference approach to confirm against your order, and no inspection outcome can be promised ahead of time.
Questions buyers ask about pre-shipment inspection
How does pre-shipment inspection differ from in-process inspection?
In-process inspection follows production as it runs, to catch drift while correction is still possible. Pre-shipment inspection looks at a completed lot and yields a release decision on it. They complement each other; the second performs badly when asked to find what the first should have caught.
How is the sampling plan for a glass bottle order decided?
The scheme supplies the mechanics, and the values are negotiated around the destination market, the contents, the filling line and the cost of one defective unit reaching a consumer. If a supplier proposes the same figure for every order, the risk level was never discussed.
How many defects are allowed in a lot?
No single figure exists without two inputs: the sample size and the agreed limit for each class. What can be fixed beforehand is the logic, with critical findings at effectively zero tolerance, major limits tightening as downstream cost climbs, and minor limits looser. State the limit per class in the protocol before production, not after counting the sample.
What should I do first when the result is a fail?
Go to the routes already named in the protocol, weigh sorting cost against downstream cost, and record the decision along with the cost bearer. Act while the goods are at the plant and the balance is outstanding, when all three routes are still available.
Which documents should travel with the inspection report?
The specification used as the yardstick, the boundary-case reference photographs, the in-process check records for that lot, and the calibration records for the instruments. Questions about whether a test report or certificate is genuine are a separate document matter.