Cannabis Sampling Plans for Batch Testing: A 2026 LP Guide

A sampling plan exists to answer one question a buyer will eventually ask: do the few grams that reach the instrument speak for the whole lot? Cannabis flower makes that hard. Bud size, trichome density and moisture vary from the top of a tote to the bottom and from one grow room to the next, so a single grab sample can miss the lot average by a wide margin.

What Health Canada requires

Health Canada does not publish a cannabis-specific container formula. Under the Cannabis Regulations, testing must be done on each lot or batch (section 91) using validated methods (section 90), and you must keep a retained sample large enough to repeat that testing. The Good Production Practices guide points licence holders to pharmacopoeial sampling procedures and names the British Pharmacopoeia's herbal drugs sampling text as an example. The British text reproduces the European one, so Health Canada is already steering you toward Ph. Eur. 2.8.20. Most LPs just haven't formalised it.

What EU importers and auditors expect

EU GMP Annex 7 states that herbal substances are heterogeneous by nature and must be sampled with special care by staff with specific expertise. If your testing lab also pulls the samples, ISO/IEC 17025 clause 7.3 requires it to hold a documented sampling plan and method. An importer's Qualified Person reviewing your file will look for both: a named method and evidence that trained staff followed it.

If your release sampling plan was written for the Canadian market alone, it isn't fit for export. Rewrite it before the next EU shipment.

Ph. Eur. 2.8.20 sets the minimum sampling procedure for herbal drugs, and dried cannabis flower falls under the herbal drug rules that monograph 3028 builds on. It works in three stages.

Stage 1: pick the containers

If the containers, labels and markings suggest the lot is homogeneous, sample all containers when there are one to three. Above three, sample the square root of the container count plus one, rounded up to the next whole number. Pick those containers at random. A lot that cannot be treated as homogeneous (two harvest rooms combined under one lot number, for example) must be split into sub-lots, and each sub-lot is sampled as its own batch.

Stage 2: take enough mass

Take one sample from each selected container, drawn from the upper, middle or lower section so the set covers different depths. For large bags, go at least 10 cm deep. The combined bulk sample must meet a minimum share of the lot's mass:

Stage 3: reduce to a test sample

Mix the bulk sample thoroughly, then reduce it by quartering: spread it in a flat square heap, divide it diagonally into four, keep two opposite quarters and remix. Repeat until you reach the minimum test sample. For flowers, that minimum is 250 g, or the whole bulk sample if it weighs less. The chapter then calls for milling through a 1 mm screen before analysis.

Run the worked example. Eighty bags means sampling 10 (the square root of 80 is 8.9, plus one is 9.9, rounded up to 10). A 40 kg lot needs 400 g, so each bag gives 40 g. One round of quartering would cut 400 g to 200 g, which is under the 250 g floor, so the whole bulk sample goes forward or you use a riffle divider that retains at least 250 g. That is 1% of the lot consumed by one release test. Budget for it in your yield.

The chapter also allows other procedures if you can demonstrate they produce representative samples. Use that clause to validate a leaner plan with data. An improvised shortcut will fail the first audit.

Ph. Eur. monograph 3028 took effect on 1 July 2024. A week later, Poland's medicines registration office (URPL) told holders of cannabis raw material registrations to update their documentation to match it. Regulators moved quickly, and EU importers now test against the monograph on arrival. When the importer's number doesn't match yours, the cause is often upstream of the lab bench. Five sampling faults are worth checking first:

• Merging harvests from different rooms into one lot number, then sampling it as if it were uniform.
• Pulling every increment from the top of each tote, where larger buds tend to collect as the tote settles.
• Reducing the bulk sample by grabbing a handful instead of quartering or using a divider.
• Leaving the sample unsealed between collection and analysis, so moisture and the dried-drug calculation shift.
• Sampling for the Canadian COA from one set of bags while the EU lab samples a different set on arrival, with no shared method.

The fourth fault matters more than it looks. Monograph 3028 reports cannabinoids on the dried drug, so a sample that lost or gained water in transit changes the reported THC figure. Our guide to water activity and moisture limits for export covers storage, and the breakdown of why EU retest numbers differ from your COA explains the calculation.

The last fault is the expensive one. Two labs can both be accurate and still disagree if they sampled differently. A shared, written method is the cheapest insurance an exporting LP can buy.

Use one plan for Canadian release and EU export so the two data sets are comparable. We call it the 5-Point Export Sampling File, and it's what an importer's quality unit should receive before the first shipment.

The 5-Point Export Sampling File

1. Define the lot. State what makes a lot homogeneous: one cultivar, one harvest room, one drying run. Anything broader becomes sub-lots.
2. Fix the container count. Cite Ph. Eur. 2.8.20 and show the square-root-plus-one calculation for your standard pack sizes.
3. Record increment mass and depth. Log the grams taken per container and where in the container each increment came from.
4. Control reduction and custody. Name the reduction method, the sealed container used and every hand the sample passed through before the lab.
5. Split and retain. Divide the reduced sample into test, retained and importer-reference portions, labelled with the lot number.

Point 5 connects to your Canadian obligations. The retained portion satisfies Health Canada's retention requirement and gives you material for a confirmatory test if an EU result is disputed. Our article on retention samples under Annex 19 sets out how much to hold and for how long. Confirm the testing lab's accreditation scope covers the sampling work too. The ISO/IEC 17025 verification guide lists what to check.

Send the file with your first COA. A Qualified Person certifying the batch can only rely on your data if it can see how the sample was drawn.

AlphaLeaf is a Health Canada Licensed Producer based in Montreal, growing hand-trimmed flower indoors from refined genetics. Every lot we ship carries ISO/IEC 17025 batch testing and full traceability from harvest room to pallet, with export authorisation under the Cannabis Act and certifications that EU import partners can verify.

Representative sampling decides whether a COA holds up in Europe. The lab result is half the file. If you're qualifying a Canadian supplier for EU import, talk to our team about our sampling and release documentation.

How many containers do you sample from a cannabis batch under Ph. Eur. 2.8.20?

Sample every container when the batch has one to three. Above three, sample the square root of the container count plus one, rounded up to the next whole number, chosen at random. An 80-bag lot needs 10 bags sampled, and a 100-bag lot needs 11.

How much cannabis flower is needed for a representative bulk sample?

Ph. Eur. 2.8.20 sets the bulk sample as a share of lot mass: 1.00% for lots under 50 kg (minimum 125 g), 0.50% for 50 to 100 kg, and smaller shares for larger lots. The test sample for flowers must be at least 250 g, or the whole bulk sample if it weighs less.

Does Health Canada require a specific sampling plan for cannabis testing?

Health Canada requires testing of each lot or batch with validated methods and a retained sample large enough to repeat that testing, but it does not publish a cannabis container formula. Its Good Production Practices guide points licence holders to pharmacopoeial herbal drug sampling procedures.

Can poor sampling cause an EU batch release failure?

Yes. If an importer's laboratory samples different containers or handles moisture differently, its cannabinoid result can fall outside the labelled tolerance in monograph 3028 even when your own COA passed. A Qualified Person cannot certify a batch that fails the importer's release testing.

Can an LP use a smaller sample than Ph. Eur. 2.8.20 prescribes?

The chapter allows other procedures if they are shown to produce representative batch samples. That means validating the alternative with data, documenting it in an SOP and sharing it with the importer's quality unit before shipment.