A pre-shipment inspection certificate is one of the most trusted documents in agricultural trade and one of the most misread. It certifies that a sample met a specification. Whether that tells you anything useful about the 25 tonnes in the container depends entirely on how the sample was drawn, and that part is usually the least discussed line in the whole contract.
This article is mostly about sampling, because sampling is where the money is.
The number that should change how you think about this
Take aflatoxin, since it is the parameter most likely to reject an Indian consignment and the one with the most carefully written sampling law behind it.
Under the European sampling rules for groundnuts, other oilseeds, apricot kernels and tree nuts, a lot of 15 tonnes or more is sampled like this. One hundred incremental samples, each of about 200 grams, are taken across the lot. That gives an aggregate sample of 20 kg, which is mixed and divided into two equal laboratory samples of 10 kg, and each of those is separately ground finely and mixed thoroughly to achieve complete homogenisation before any analysis happens.
Large lots are not treated as one lot. A consignment of 500 tonnes or more is divided into sublots of 100 tonnes, and each sublot gets its own 100 increments and its own 20 kg aggregate sample. So a 500 tonne lot means five sublots, 100 kg of aggregate sample in total and ten laboratory samples. Below 15 tonnes the plan scales down, with a minimum of 10 and a maximum of 100 increments and an aggregate from 2 kg upward, and division into two laboratory samples only once the aggregate reaches 12 kg.
Now compare that with what commonly happens commercially. A surveyor visits a warehouse, opens some accessible bags, takes a few handfuls, composites perhaps a kilogram, and sends it for testing. The laboratory does careful work on what arrives. The certificate says the lot passed.
Those two procedures are not different in degree. They are answering different questions.
Why the plan is that large
The regulation states the reason directly. Sampling plays a crucial part in the precision of the determination of mycotoxin levels, which are very heterogeneously distributed in a lot. For products with a large particle size, groundnuts and figs being the named examples, the distribution is even more uneven, which is why the aggregate sample weight is increased for those products specifically to keep representativeness equivalent.
The physical picture behind that is worth holding onto. Aflatoxin does not spread evenly through a lot like salt in water. It concentrates in a small number of badly affected kernels. A lot can be well within the limit on average while containing individual kernels at thousands of times the limit. Whether your sample happens to catch one of them is close to a lottery if you take a small sample, and becomes a reliable measurement only when you take many small increments from many places and grind the whole thing together.
Which produces the conclusion that matters commercially: for mycotoxins, the sampling step contributes more error than the laboratory step. Two competent laboratories reporting different numbers on the same container are usually both right about the samples they received. The disagreement was created before either of them opened a box. The limits themselves, and why the same lot can be lawful in one market and not another, are in the aflatoxin article.
Three practical consequences:
- Arguing about laboratories when results diverge is usually the wrong argument. Ask how each sample was drawn.
- A tighter contractual limit does not compensate for weak sampling. It just moves where the lottery pays out.
- Where the parameter is heterogeneous, paying for a bigger sample is better value than paying for a better laboratory.
The same logic applies with less force to other parameters. Moisture and foreign matter are more evenly distributed, so smaller samples give more stable answers. Pesticide residues sit in between. Pathogens are the extreme case, distributed even more unevenly than mycotoxins, which is why sampling design matters more than analytical method for Salmonella in a low-moisture food.
What is actually being inspected
A useful pre-shipment inspection covers four things, and inspections that go wrong usually covered fewer.
| Object of inspection | What is checked | Common gap |
|---|---|---|
| The lot | Sampling, grading, laboratory parameters against the contract | Sample not representative, or drawn from a different lot |
| The packing | Bag type, weight, stitching, liner, markings, labelling, batch coding | Marking checked against the invoice rather than the buyer's artwork |
| The container | Cleanliness, dryness, previous cargo, odour, light test, floor condition | Not inspected at all, because it arrives after the surveyor leaves |
| The documents and the link | Lot identity, quantity, seal number, photographs, dates | Nothing ties the certificate to the container that actually sailed |
That last point deserves stating plainly. An inspection report with no container number and no seal number is an inspection of a warehouse, not of a shipment. It cannot answer the only question that matters in a dispute, which is whether the goods inspected are the goods delivered.
Sampling procedure, in the order it happens
- Define the lot. A lot has to be a defined, homogeneous, identifiable quantity in one place. Sampling across two production batches and calling the result one lot is a common and serious error.
- Take incremental samples across the lot, not from accessible points. That means bags from the back and the middle of the stack, several depths within a bag, and a spread across the full quantity. From a moving stream, at defined intervals over the whole run.
- Composite into an aggregate sample. The number of increments and the aggregate weight should follow a named plan, not a habit.
- Reduce the aggregate correctly. Riffle division or quartering, not scooping off the top. Reduction is where a representative aggregate quietly becomes an unrepresentative laboratory sample.
- Grind and homogenise where the plan requires it. For a large-particle commodity with heterogeneous contamination this step is not optional, and skipping it makes the aggregate size pointless.
- Split into laboratory sample, retained samples and the buyer's sample. Seal each, label each, photograph the seals.
- Record everything. Date, place, lot, quantity, method, who drew it, seal numbers.
Step two is where most of the loss happens, and it is invisible in the report. Ask your surveyor to describe how increments were distributed across the stack, and ask for photographs of the stack rather than of the sample bag. A surveyor who took increments properly is happy to describe it.
Timing, and why supervised stuffing wins
There are three moments an inspection can happen, and they answer different questions.
Before stuffing. Sampling and testing on the finished, packed lot in the warehouse. Good for the laboratory parameters, because the results are available before the container is committed. Says nothing about the container or about what was actually loaded.
During stuffing, supervised. The surveyor watches the container inspection, the loading, the count and the sealing, and draws samples from the flow. This is the strongest option, because it is the only point at which the lot, the packing, the container and the seal are all witnessed together and tied to one document.
After stuffing or on arrival. Answers the question of what arrived, which is what a buyer actually cares about, but too late to prevent anything and after transit has had its effect.
The practical arrangement that works is testing before stuffing so the results are known in time, plus supervised stuffing so the tested lot and the shipped container are demonstrably the same, plus a sealed retained sample so the position can be reconstructed later. Container-specific checks, the six-sided inspection, the light test, the previous-cargo question and moisture management, are covered in the container loading guide.
Independence, and who pays
Somebody has to appoint the surveyor and somebody has to pay, and both create an appearance of influence. There is no arrangement that eliminates it, so the workable answers are structural.
- Name the surveyor in the contract, before the shipment and before any dispute. Both parties agreeing on a name in advance is worth more than either party's assurances about independence afterwards.
- Allocate the cost in the contract as well, including who pays for re-inspection and re-testing if the first result fails.
- Use an accredited laboratory and require the accreditation scope to cover the specific method. Accreditation for one matrix does not cover another.
- Require the method and the limit of quantification on the report, not a bare pass. A pass against an unstated limit of quantification is not a result, and this comes up constantly with residues, where a laboratory reporting "not detected" at a high limit of quantification is reporting its own insensitivity.
- For a US buyer, remember that your own audit does not count. Sampling and testing is the one verification activity a foreign supplier is permitted to carry out itself under the American rules. The detail is in the FSVP article.
What the report should contain
A report that has all of that is a document you can defend two years later. A report that says "quality found good, packing found sound" is a courtesy note.
Retained samples, the cheapest insurance in the trade
A sealed retained sample costs almost nothing and settles arguments that would otherwise be unsettleable. Draw at least three at the time of sampling, seal and number them, and hold one with the exporter, send one to the buyer and keep one with the surveyor or in a neutral place. Agree how long they are kept, and agree that a dispute is assessed against the retained sample rather than against adjectives.
This is particularly powerful for the parameters that cannot be reduced to a number. Colour, eye definition in a black-eyed cowpea, the visual grade of a dal, the specific look of a dehusked kernel. Those are all judgements, and a judgement is only enforceable against a physical reference. It is why the retained sample turns up in nearly every specification article on this site.
Load port and discharge port will not agree
They will not, and a contract that assumes they will is a contract with a fight in it.
| Parameter | What transit does | Contractual handling |
|---|---|---|
| Moisture | Migrates within the container, condenses, redistributes | Name load port or discharge port as governing, and a tolerance |
| Insect activity | Eggs present at load hatch during a warm voyage | Fumigation record, and a live-insect clause with a defined remedy |
| Colour | Fades with time, light and heat | Measure at load, name the date, expect arrival to be lower |
| Mould and caking | Develops from a moisture problem that existed at load | Moisture control plus liner and desiccant specification |
| Aflatoxin | Does not increase without moisture, and can increase with it | Moisture control is the aflatoxin control in transit |
| Broken and split | Increases with handling and transfer | Measure at load, allow a handling tolerance |
The last row of that table is the general principle. Almost every difference between a load-port and a discharge-port result has a physical explanation that has nothing to do with anyone's honesty. Deciding in advance which measurement governs, and what tolerance applies, converts an argument about integrity into an arithmetic question.
How we handle it
We treat sampling as the part of the inspection worth spending money on. Increments are spread across the stack rather than drawn from what is accessible, laboratory samples are reduced and homogenised rather than scooped, and reports carry the method and the limit of quantification rather than a bare pass. Stuffing is supervised so the tested lot and the sealed container are tied to the same document with the seal number recorded, and retained samples are drawn, sealed and numbered at the time of sampling with a copy for each side.
Where a buyer nominates their own surveyor and laboratory, we prefer it. An independent result we did not arrange is worth more to the relationship than a favourable one we did.
The documents that carry this into the shipment are in the export documentation guide, the plant health side is in the phytosanitary and fumigation article, and the packing decisions the inspection has to verify are in the packaging article.