Sampling Botanical Ingredients: Getting a Sample That Represents the Batch
Every certificate of analysis makes the same quiet claim: that the material in the laboratory reflects the material in the warehouse. That claim is verified far less often than it is assumed. A result describes the sample that was tested, and the sample is normally drawn by a person with a scoop, under time pressure, from a drum that has been shaken for three weeks in a container. Where a dispute arises over a botanical lot, the sample — not the instrument — is the most common point of failure. This guide sets out how to draw a sample that can carry the weight of a release decision, which samples a lot actually needs, and where a properly drawn sample gets ruined before it reaches the bench.
What a sample has to survive to be called representative
The chain from a delivered lot to a reported figure has four links: the lot, the increments drawn from it, the composite built from those increments, and the subsample taken in the laboratory for the test portion. Representative sampling means that error is controlled at every link, and the error introduced by sampling is generally larger than the error introduced by the analysis itself.
The first question is therefore not how to sample but what decision the sample has to support. If the decision is whether to accept or reject a twenty-drum lot, the sample has to span the lot, which means increments from several drums rather than a generous sample from one. If the decision is whether drum seven can be released to production after a moisture excursion in the warehouse, then drum seven is the decision unit and sampling the rest of the lot adds nothing. A material with a demonstrated history of homogeneity within a lot needs fewer increments than one that is blended or repacked by a third party, and a supplier’s lot definition is part of that assessment.
Sampling from drums, sacks and bulk bags
Powders separate. Fine particles migrate downwards and coarser material stays nearer the surface, and the vibration of transport accelerates the process, so the top of a drum is not a smaller version of the drum. A sample drawn from the surface layer of a settled powder can differ from the bulk in particle size, in bulk density, in colour and, because the assay is measured on a test portion, in apparent content.
The practical response is to draw increments through the depth of the container rather than from its surface, using a sampling spear or thief that reaches the base of the layer being assessed, and to take several increments from different positions across the consignment. Where the material can be sampled from a moving stream — during transfer, packing or blending — that is the strongest option available, because a moving stream is more uniform than a settled bed and a set of increments taken across the whole stream is closer to the ideal of sampling every part of the lot. Bulk bags should be sampled at more than one height, since the spout alone draws from a single location.
Part-used containers deserve separate treatment. Once a drum has been opened, the head space becomes the wettest and most reactive part of the container, and the material at the surface can be measurably different from the material below it. Where a container has been opened and stored, sampling the surface layer as though it represented the drum is a reliable way to generate a moisture or microbial finding that the rest of the drum would not reproduce. Taking the top layer as its own sample, and the remainder as another, tells you more than either alone.
The tools matter more than they are usually given credit for. A sampling instrument should be clean, dry, non-shedding and dedicated to the material, because a corroded spear contributes metal, a damp scoop changes the moisture result and a shared tool can carry an allergen from the previous product. Oils, pastes and semi-solid materials have to be brought to a workable condition and mixed before an increment is drawn, and they are the materials where a single dip from the surface most reliably misses the answer.
Composite, individual and retain samples answer different questions
A composite sample is built by combining increments and testing the mixture once. It is the routine release sample, and it is efficient because it produces one result for a lot. Its limitation is built into its virtue: blending the increments erases the difference between them, so a composite can never answer a question about within-lot variation. Where the question is whether the lot is homogeneous, or whether one drum is different, the units have to be sampled and tested individually.
The retain sample is a different object again. It is not consumed by testing; it is sealed, labelled and stored so that a later question can be answered against the material as it was when it was received. It is what makes a dispute evidential, and it is what allows a comparison with the material that precedes or follows a change. Retain samples should be held under the storage conditions the specification claims, because a retain sample that has taken up moisture in a humid store cannot be used to argue about the moisture of the lot.
Microbiological samples must be drawn separately and aseptically, into a sterile container, with a sterile tool. A sample taken from the composite is not a microbiological sample, because the composite was exposed, the scoop was not sterile and the result describes the handling as much as the material. The same separation applies to a moisture sample, which should be sealed immediately with as little head space as possible, and to samples intended for authenticity or adulteration testing, where an unbroken chain of custody from container to laboratory matters because the finding may be contested. Ageing references in a dispute is far more expensive than documenting the draw at the time.
Subsampling: where a good sample is ruined
A correctly drawn composite can still produce a wrong result if the laboratory’s subsampling is careless. If the powder segregates, splitting the composite with a spoon produces a portion that is not representative — one portion richer in fines, another richer in coarse particles, and different assay results from the two. A riffle or rotary splitter, used repeatedly, is the standard way to reduce a sample without changing its composition, and coning and quartering serves for coarser materials where a splitter is impractical.
Grinding before analysis is another point at which the material is changed rather than simply divided. Milling generates heat, which can drive off volatiles and shift a moisture result; it changes the particle size distribution, which changes how completely the assay solvent extracts the marker; and it can blend a segregated sample into temporary uniformity that the test portion then reports as the truth. Where a method specifies grinding, the mill, the duration and the resulting particle size belong in the method statement.
The size of the subsample is a function of the method rather than of convenience. A test portion for a marker present at a fraction of a percent requires more material than one for a marker present at tens of percent, because the quantity being detected is smaller and the sampling error is correspondingly larger. Where a method’s sensitivity and the subsample size are not matched, the laboratory reports a number with a precision the sample cannot support.
Why two laboratories can disagree about the same lot
When two laboratories report different values for one lot, four explanations account for almost all cases, and they are worth ruling out in a defined order: the sample was not the same; the subsampling was not equivalent; the method differed; or the calculation basis differed, such as a result expressed on a dried basis in one report and as-received in the other.
The order matters because the explanations are not equally likely and not equally cheap to test. Method differences are genuinely common in parameters such as particle size and residual solvent, where instrumental technique changes the answer. But a fresh, correctly drawn sample split between the two laboratories eliminates the first two explanations at very little cost, and it is the step most often skipped in favour of an argument about instruments. Where an assay result is disputed, sending a split of a single homogenised sample to both laboratories is the fastest route to knowing which party has the problem.
Writing the sampling procedure down
A sampling plan that exists only in an experienced colleague’s habits is a single point of failure, and it will fail at the moment it is needed most. What belongs in writing is modest: who is authorised to sample, which tool is used, which containers the samples go into, how the samples are labelled — product, lot, container identity, date, sampler and sampling position — and how many increments are drawn per container and per lot.
The increment count is the parameter that most needs a rule rather than an instinct. The general principle is that the number of increments should increase with lot size and with the amount of variation expected, and a plan can be designed from a known or estimated variation rather than from a convention. Where the history of a material shows tight uniformity, a leaner plan is defensible; where a material is blended, repacked or supplied across several sub-lots, a heavier plan is the honest choice. The plan should also state the sample size, the retention period, the storage conditions for the retain, and the circumstances in which chain-of-custody documentation is raised.
Finally, the plan should be reviewed when the logistics change rather than when a failure occurs. A procedure written for twenty-five kilogram sacks does not sample a five-hundred kilogram bulk bag properly, and a plan designed for single-site deliveries does not cover a shipment that has been consolidated from several production runs.
What to write into the purchase specification and the quality agreement
State who draws the release sample and who holds the retain sample, and for how long. State the sampling method, the number of increments and the sample size, so that a result can be reproduced by a second party. State the definition of a lot, because a lot number that covers several blends or several production days is a claim about uniformity that has to be justified. State that the certificate must identify the sampling method and the lot it describes, alongside the material’s identity and the specification revision it was measured against. And state what happens when a result is disputed: the sample to be compared, the laboratory, and the point at which a fresh sample is drawn rather than the original argued over.
Frequently Asked Questions
How many increments do I need from a consignment?
There is no single correct number, because it depends on the size of the lot and on how uniform the material is known to be. The principle is that the increment count rises with lot size and with expected variation, and that the sample must span the lot that the decision covers. Set the plan against a known or estimated variation, review it when packaging or suppliers change, and record it so that the same plan is applied every time.
Can I use one sample for every test?
Not for microbiological testing, which requires a separate aseptic sample into a sterile container with a sterile tool, and not for a retain sample, which is kept sealed rather than consumed. Moisture is best sampled and sealed separately, and authenticity testing may require documented chain of custody. A composite is the right sample for routine chemistry and the wrong sample for a question about variation within the lot.
Why do two laboratories disagree about the same lot?
In order of likelihood: the samples were not the same, the subsampling differed, the methods differed, or the results were expressed on different bases. Rule out the sample first by splitting a freshly drawn, homogenised sample between the two laboratories. Where the method is genuinely the cause, as it often is for particle size and residual solvent, the method has to be named in the specification.
How long should a retain sample be kept?
Long enough to cover the period in which a question can reasonably be asked, which is usually tied to the shelf life of the material and to the retention requirement of your market and your customers. Keep it under the storage conditions stated in the specification, and keep the records that identify which lot and which container it came from, because a retain sample without provenance cannot settle anything.
Does sampling matter if the powder is homogeneous?
It matters most for the materials that are described as homogeneous. Blends, standardised extracts and repacked powders are the ones where an assumption of uniformity is made rather than demonstrated, and where segregation during transport produces real differences between the top and the bottom of a container. Demonstrating homogeneity once, with individual unit samples, is what justifies a leaner routine plan afterwards.
