Any laboratory that presents its results as absolute truth is overselling. Measurement carries uncertainty, and cannabis testing carries more of it than most analytical fields — for reasons that are worth understanding rather than glossing over.
None of what follows is an argument against testing. It is an argument for reading results as measurements with a margin, which is what they are.
Sampling Is the Largest Source of Variability
The laboratory does not analyse your batch. It analyses a sample, and the sample is where most of the variation originates — usually more than the analysis itself.
Cannabis is heterogeneous at every scale. Cannabinoid concentration varies between the top and bottom of a single plant, between plants in the same room, and between containers in the same harvest lot. Microbial contamination is frequently localised rather than evenly distributed, so a compliant composite can coexist with a genuinely contaminated pocket somewhere in the lot.
Sampling protocols exist to manage this, and they help substantially. They do not eliminate it. A representative sample is an approximation of the batch, and the approximation gets rougher as the batch gets larger and less uniform.
Why Two Laboratories Disagree
Send material from one batch to two accredited laboratories and you will often get different numbers. Several legitimate factors contribute.
Sample preparation. Grind size, extraction solvent, extraction time, dilution — all affect how much of the target compound reaches the instrument.
Calibration. Different reference standards, different calibration ranges, different curve fits.
Peak integration. Chromatographic peaks require judgement about where they begin and end. Reasonable analysts make slightly different calls, and those calls move numbers.
Moisture basis. Potency reported on a dry-weight basis differs from as-received. If two reports use different bases, the figures are not comparable even when both are correct.
Instrument condition. Detector response drifts between services and recalibrations.
Individually these are small. Stacked, they explain most of the divergence people attribute to bad faith. Where bad faith does enter, it is usually through the incentive structure described in lab ethics and data integrity — but incompetence and honest variation account for far more of the spread than fraud does.
Where Methods Genuinely Struggle
Complex matrices
Methods validated on flower do not transfer cleanly to chocolate, gummies or beverages. High-fat matrices bind cannabinoids and resist extraction; sugars and colourants interfere with detection. Every matrix needs its own validation, and the industry has more product formats than it has validated methods. See edibles testing.
Pesticide lists that do not match reality
Panels screen for a defined list of compounds. A pesticide not on the list is not detected, however much of it is present. Lists are set by regulation and updated slowly, while the compounds actually in circulation change faster. A clean pesticide result means clean for the compounds screened.
Microbial method disagreement
Culture-based methods detect viable organisms. qPCR detects genetic material, including from organisms that are no longer alive. The two answer different questions and can disagree on the same sample, and jurisdictions differ on which they accept. Neither is wrong; they are measuring different things.
Emerging contaminants
Standard panels were designed around known risks. Novel synthetic cannabinoids, unregulated additives and newly identified degradation products are not necessarily covered. The EVALI outbreak involved vitamin E acetate, which no routine panel was screening for at the time — that is the general shape of the problem.
Total THC calculations
Total THC is calculated from THC and THCA using a conversion factor for decarboxylation. The factor assumes complete conversion, which real-world heating does not achieve. The reported figure is therefore a theoretical maximum rather than what a consumer receives. More in THC percentage explained.
What Results Cannot Address At All
Some limitations are not analytical. They are outside the scope of any laboratory.
Post-testing handling. A certificate describes the sample on the date of analysis. Heat, light, humidity and time change products afterwards, and no result accounts for a warehouse that ran warm in August.
Individual response. Potency is chemical concentration. It does not predict how a specific person will respond — tolerance, physiology, route of administration and context all matter, and none of them appear on a COA.
Quality in the broad sense. Passing means meeting limits. It does not describe how well something was grown, cured or made.
What Is Improving
Several of these problems are being worked on, with real if uneven progress.
Standardised methods reduce inter-laboratory variation by narrowing the range of defensible choices. Blind proficiency testing gives regulators a way to detect systematic drift rather than relying on self-reporting. Independent sampling removes the easiest route to an unrepresentative sample. Improved instrument sensitivity is lowering detection limits, and better matrix-specific validation is slowly catching up with the product formats on the market. Where that is heading is covered in standardisation and transparent reporting and the future of cannabis lab testing.
Working With Uncertainty
The practical response is not to distrust results. It is to use them properly.
- Treat potency as a range, not a point. If a laboratory can tell you its measurement uncertainty, that number is more informative than the headline figure.
- Track trends rather than single results. Ten batches from one laboratory tell you far more than one batch from three.
- Stay with one laboratory for comparability. Consistent methodology makes your own data set meaningful, and moving after an unwelcome result is the behaviour the incentive problem depends on.
- Read the whole report. Detection limits, methods and dates carry information the headline numbers do not.
- Do not read a pass as a guarantee. It is a measurement against a threshold, with a margin.
Honest laboratories are comfortable discussing all of this. A laboratory unwilling to acknowledge that its results carry uncertainty is telling you something about how it handles the results themselves.