Sampling Matters: What One Hemp Test Can—and Cannot—Represent

A laboratory analyzes the material it receives. Whether that result can describe a plant, field, production batch, bottle, jar, or individual unit depends on what was sampled and how.

The instrument may perform perfectly while the conclusion is still too broad. Sampling is the bridge between a test portion and the larger material someone wants to understand.

The chain from lot to result

A typical sequence is:

defined lot → selected specimens → composite or representative sample → laboratory sample → test portion → reported result

Each step narrows the physical material. The final number comes from the test portion, not from direct measurement of every item in the lot.

Five terms to keep separate

Term	Practical meaning
Lot or batch	The defined quantity the sampling plan intends to address
Increment or specimen	Material selected from one place or unit
Composite sample	Material formed by combining selected increments
Laboratory sample	Material received and identified by the laboratory
Test portion or specimen	The smaller prepared amount actually analyzed

Programs and laboratories may use different terminology. The report, sampling plan, and chain-of-custody record should explain their identifiers.

Define the lot first

Representative sampling begins with a defined population.

For USDA crop compliance, a “lot” refers to hemp crop acreage identified by the producer under the production program. The sampling protocol selects plant specimens intended to represent the total THC concentration of that defined lot.

A retail production batch is different. It might be a quantity of flower packaged together, a tank of oil, a gummy production run, or a lot of isolate. The manufacturing record should define its boundaries.

If material from different harvests, tanks, dates, formulas, or suppliers is combined without a documented lot definition, one test cannot repair the missing traceability.

Selection must address variability

Hemp material is not necessarily uniform. Variation can occur:

• among plants;
• within a plant;
• across a field or room;
• between harvest dates;
• among flower sizes or positions;
• between portions of bulk powder;
• across a mixing tank;
• between finished units; and
• during storage or handling.

A sampling plan should consider the expected sources of variation. Taking all material from the top of one container may be convenient, but it may not represent material that settled or separated.

More material is not automatically more representative. The locations, timing, randomization or systematic selection, and handling can matter more than the gross sample weight.

Representative does not mean identical

A representative sample is selected through a design intended to support an inference about a defined population. It is not a miniature copy in which every property and every rare event appears.

The strength of the inference depends on factors such as:

• how clearly the lot was defined;
• how selection locations or units were chosen;
• the number of increments;
• expected variability;
• whether selection bias was controlled;
• whether the sample remained intact; and
• whether the analytical test portion was prepared consistently.

A sampling plan can support a stated level of confidence for a particular decision without guaranteeing every unit. The confidence statement belongs to that plan, population, and decision—not to all future lots.

Composite samples combine information—and boundaries

A composite sample combines increments before analysis. It can estimate an average for the defined lot when the design is appropriate.

That average may hide local variation. A compliant composite result does not prove that every individual plant or gummy equals the reported concentration.

Conversely, testing a single unusual unit does not necessarily describe the batch average. The appropriate design depends on the question:

• Is the goal to estimate a lot average?
• Detect a rare contaminant?
• Confirm unit-to-unit uniformity?
• Identify a worst-case location?
• Make a regulatory pass/fail decision?

Different questions can require different plans.

USDA crop sampling has a specific scope

USDA’s sampling guidelines are designed for hemp grown under a production plan. Trained sampling agents collect specimens within the required timing before anticipated harvest. The specimens are intended to represent a homogeneous lot for total THC determination.

At the laboratory, USDA guidance calls for drying, grinding the composite plant sample, separating test and retain specimens, and reporting total THC on a dry-weight basis with measurement uncertainty.

That system answers a production-compliance question about a crop lot. It does not automatically establish:

• finished flower-package uniformity;
• potency of an extract made later;
• cannabinoid amount per gummy;
• pesticide or metals status unless those panels were separately tested;
• stability months after sampling; or
• legality under every retail-product rule.

Homogenization changes the question

Grinding and mixing can improve the consistency of a laboratory test portion. If a plant sample is ground thoroughly, small portions drawn from the prepared material may be more similar to one another.

Homogenization does not make the original field or production batch perfectly uniform. It makes the prepared laboratory material more suitable for measuring an average under the method.

For finished products, the laboratory may homogenize several gummies or a portion of oil. The report should say what material was combined. A homogenized multi-unit result cannot automatically verify the amount in each individual unit.

Contaminants can require different thinking

Cannabinoid concentration and contamination may have different distributions.

A cannabinoid may be broadly distributed in a well-mixed oil. A microbial problem might be localized. A foreign particle or pesticide residue might occur in only part of a lot.

A sampling design that estimates average potency may not have the same probability of finding a rare, unevenly distributed contaminant. Read the regulator’s or method’s sampling requirements for the specific panel.

Do not infer that a potency sample represents every contaminant question.

Sample handling matters after selection

The selected material can change or become compromised through:

• heat;
• light;
• moisture gain or loss;
• contamination;
• damaged seals;
• long transit;
• incorrect containers;
• incomplete labeling; or
• broken chain of custody.

The record should show who collected the sample, when, from which lot, how it was sealed, how it was transported, when the laboratory received it, and whether its condition was acceptable.

Batch, Lot, Sample ID, and Test Date explains how to connect those records.

Retain samples and retesting

A laboratory may keep a retain specimen for reanalysis or confirmation. Retesting the retained material can answer whether the laboratory obtains a similar result from that retained sample.

It does not create a new independent field or production sample. If the original selection was unrepresentative, repeated testing of the same retained composite may reproduce the same narrow view.

Resampling returns to the lot and selects new material under an authorized protocol. Some regulatory programs restrict when and how resampling may occur to prevent selective testing until a preferred result appears.

What a COA should reveal

Look for:

• lot or batch identifier;
• sample identifier;
• matrix;
• sampling date;
• sampler or submitting party;
• collection method or regulatory program;
• sample quantity or unit count, where provided;
• laboratory receipt condition;
• preparation method;
• test and analysis dates; and
• any statement limiting the result to the submitted sample.

A laboratory may not have controlled sampling when a client mailed in material. “Client submitted” can mean the lab accurately tested what arrived but cannot attest to how it was selected from the larger batch.

Three levels of conclusion

Supported by the analysis

The laboratory reported a result for the identified test sample using the stated method.

Supported only with sampling evidence

The result is representative of the defined lot under the documented sampling plan.

Usually not supported by one result

Every individual unit contains exactly the reported amount, every location is contaminant-free, or future batches will produce the same result.

Keeping these levels separate prevents a precise result from being stretched into an unsupported guarantee.

Questions for a finished product

• Was the sample an ingredient, bulk mixture, or packaged finished product?
• How many units or locations contributed?
• Were units combined before testing?
• Was selection random, systematic, targeted, or client chosen?
• Does the result describe an average or individual unit?
• Was the tested matrix the same as the product sold?
• Did the batch change after sampling through blending, packaging, or added ingredients?
• Are the sampling and laboratory dates compatible with the product record?

For gummies or capsules, a finished-product average does not prove perfect unit uniformity. For oils, a tank sample may not describe later bottles if the mixture separated or the formula changed. For flower, a small sample does not describe a different harvest lot.

Average potency and unit uniformity

Imagine ten gummies are combined, homogenized, and tested as one composite. The result may estimate the average amount across the combined material. It cannot reveal whether each gummy contained that average.

Testing individual units can provide unit-level data, but the number and selection of units still determine how broadly the results can be generalized. A manufacturer’s uniformity study, a regulator’s compliance sample, and a consumer-submitted gummy can all be valid tests while answering different questions.

The report should not be summarized as “every gummy tested at X” unless individual-unit testing and the sampling record support that statement.

Sampling after a process change

A result belongs to the batch and process state that existed when sampling occurred. Additional blending, dilution, flavoring, heating, drying, remediation, repackaging, or combining lots can change the material.

When a later step changes composition or creates a new production batch, the earlier sample may become an ingredient or intermediate record. It should not silently follow the product as though the finished material was sampled.

Compare the sample date with manufacturing and packaging records. If the sampled stage is unclear, request the chain of custody or batch record before describing the COA as finished-product evidence.

Reader checklist for a seller-provided COA

• Is the lot on the package the lot that was sampled?
• Who selected the material—the regulator, laboratory, producer, or an unknown submitter?
• How many locations or units contributed?
• Is the sample composite or individual?
• Does the matrix match the sold format?
• Was the product changed after sampling?
• Does the report limit conclusions to the submitted sample?
• Is the claim about an average, a unit, or contaminant absence?
• Would a different sampling design be needed for that claim?

A careful summary

Instead of “the whole product line tested at 20%,” write:

The laboratory reported 20% for the identified sample from the stated batch. The extent to which that result represents the full batch depends on the documented sampling plan and batch controls.

That wording preserves the evidence and the boundary.

Primary sources

• USDA AMS, Sampling Guidelines for Hemp
• USDA AMS, Hemp Laboratory Testing Guidelines
• California DCC, Medicinal and Adult Use Cannabis Regulations, revised January 1, 2026
• NIST, CannaQAP Exercise 2