Why Potency Labels Don't Match Between Cannabis Labs

September 27, 2026

Photo by Mezidi Zineb via Unsplash.

Send the same jar of flower to five different labs and you'll get five different numbers back. Not close numbers, either. A homogeneous, spiked distillate sample -- meaning identical material, mixed thoroughly enough that variation from the plant itself is essentially eliminated -- was split and sent to five labs for standard potency testing. The results came back anywhere from 77% to 94% THC. That's not a rounding discrepancy. That's a 17-percentage-point spread on a sample that was, by design, exactly the same in every single vial.

Most growers and most buyers treat the number on a certificate of analysis the way they'd treat a molecular weight or a boiling point -- a fixed physical fact about the material. It isn't. It's the output of a chemical assay with real, measurable error, run by a lab that in many states is also a business competing for the next client's testing contract. Regulators have caught up to this reality faster than consumers have. Massachusetts pulled products off retail shelves for independent re-testing in 2026. California suspended labs over alleged potency inflation. Maryland built an entire cross-lab auditing pilot specifically because the state stopped trusting the numbers labs were reporting on their own.

This piece walks through why that spread exists even when nobody's cheating, what happens when someone is, what regulators in three states are doing about it, and -- more usefully -- how a grower or buyer should actually read a lab result once you understand how much slack is built into it.

The Test Itself Isn't as Precise as You'd Think

Testing the same cannabis sample across five different labs produced THC results ranging from 77% to 94%, highlighting significant inconsistency in lab testing standards.

Start with the study that should reset everyone's expectations: identical spiked distillate, split five ways, tested for THC using standard high-performance liquid chromatography (HPLC) protocols at each lab. The 95% confidence limit on that result set was 8.08 percentage points. In plain terms, if the true value were 85%, you'd statistically expect individual lab results to land anywhere from roughly 81% to 89% just from method noise -- and the actual observed range, 77% to 94%, ran even wider than that confidence band would predict for a single well-controlled method.

Here's the part that makes it clear this is a lab problem and not a sample problem: the same material tested via mid-infrared spectroscopy, a different analytical method, came back with a 2.3% confidence interval. Same distillate, same homogeneity, radically tighter agreement. That gap between 8.08 and 2.3 percentage points is the signature of methodology and lab execution, not plant chemistry.

Standard HPLC potency testing has more steps than people assume, and each one carries its own small error. The flower gets ground to a specific particle size before extraction -- grind too coarse and you under-extract; too fine and you risk degrading cannabinoids or including plant material that skews the matrix. Extraction efficiency depends on solvent choice, time, and temperature, and no two labs run identical protocols unless they're required to. Calibration standards have to be prepared and checked against reference material, and drift here compounds directly into the final number. Then there's moisture correction, since THC percentage is reported on a dry-weight basis and moisture content itself is a measurement with its own error margin. None of these steps is dramatic on its own. Stacked together, they explain most of that 8-point spread before anyone touches a number with intent.

The same research looked at pesticide screening and found something arguably more alarming: a 78% false negative rate. That means contaminated samples were passing as clean the large majority of the time. A THC number that's off by a few points is a labeling annoyance. A pesticide test that misses contamination three times out of four is a real safety gap, and it's a reminder that potency inflation isn't the only thing riding on lab reliability.

When It's Not Just Error -- The Inflation Problem

Photo by elsaolofsson via Pixabay.

Measurement error is one thing. What California's Department of Cannabis Control (DCC) started uncovering is another. In April 2025, the DCC suspended BelCosta Labs over allegations of THC potency inflation along with fraudulent pesticide and mold testing -- serious enough charges that the agency moved to pull the lab's license outright.

The case took over a year to resolve. On July 24, 2026, Administrative Law Judge Cindy Forman ruled that the state hadn't presented sufficient evidence of unreliable results and ordered BelCosta's permit restored on probation, characterizing the DCC's enforcement action as unduly punitive. On paper, BelCosta won. In practice, CEO Myron Ronay has said the business is closed and will likely never reopen. That's the part worth sitting with: a lab can be legally vindicated and still be commercially dead, because in a market where trust is the entire product, an accusation does damage that a favorable ruling can't undo. It's also a signal to every other lab in the state that enforcement scrutiny, even when it doesn't stick, carries real cost.

Separately, and around the same window, Verity Analytics received an April 19 suspension notice alleging THC inflation of up to 32% -- not a rounding error, a near-third overstatement -- along with allegations that a lab director was clearing pesticide samples by eyeballing chromatograms rather than applying defined scientific criteria. Eyeballing a chromatogram instead of running it against quantitative thresholds is exactly the kind of shortcut that turns a testing lab into a rubber stamp.

None of this is new pressure, either. Before California tightened its rules in 2024, it was routine to see flower labeled at 30-45% THC, numbers that cannabinoid chemists will tell you are close to biologically implausible for cannabis flower grown under normal conditions. The underlying incentive is simple and structural: labs compete for cultivator and brand clients, and in a market where consumers shop by percentage, the lab willing to report the highest number tends to win the next batch of samples. That's not a conspiracy, it's just what happens when a testing industry gets paid by the same people whose product it's grading.

States Are Starting to Check the Checkers

Photo by hin via Pexels.

California's response was to stop treating THC testing as something each lab could interpret its own way. Since January 1, 2024, the DCC has required all licensed labs to follow standardized operating procedures specifically for measuring THC potency, closing off a lot of the method-to-method variance that let inflated numbers hide behind analytical technicalities. It's a blunt fix, but a necessary one -- you can't compare labs fairly if they're not running the same test.

Massachusetts took a more active enforcement approach. In May 2026, the Cannabis Control Commission announced it would start pulling retail products directly off shelves and independently re-testing their potency rather than relying solely on the certificate the product shipped with. By August 2026, the CCC had re-tested 63 flower products from 62 different producers, checking results against a permitted range of 75% to 125% of the labeled potency. Thirteen of those products fell outside that window and were placed on hold. That's roughly one in five products failing an independent check -- a meaningful failure rate for an industry that had, until recently, operated almost entirely on self-reported numbers.

Maryland's approach, launching as the CSQ CLEAR pilot in October 2026, goes further structurally. Instead of one regulator spot-checking products, the program cross-tests retail flower samples across four independent labs, all benchmarked against the state's official reference facility. That design solves the exact problem the five-lab distillate study exposed -- if you only ever check one lab against itself, you never see the spread. Maryland's program also applies forensic statistical modeling to flag retail products whose results don't match their original Certificate of Analysis, catching drift or inflation after the product has already reached a shelf.

The pattern across all three states is the same shift in posture: regulators have stopped assuming a lab's number is correct by default and started building the infrastructure to check the checkers. That's a real structural change, not a messaging update, and it's likely to spread to other state markets once the auditing playbooks are proven out.

What the Real Numbers Look Like When You Strip Out Inflation

Photo by Colin redwood via Unsplash.

Strip the incentive to round up out of the picture entirely and look at what peer-reviewed research actually found in tested flower. A study published in PLOS One, working with Colorado market samples, found that observed THC potency averaged 23.1% lower than the lowest label-reported values on the same products. Not lower than the average label claim -- lower than the floor of what labels were claiming. About 70% of the samples tested came in more than 15% lower than their labeled THC.

That's a meaningful gap for anyone actually trying to evaluate genetics rather than shop by number. If you're a grower comparing strains, or a buyer trying to figure out whether a cultivar is genuinely potent or just well-marketed, the label percentage has been doing a poor job of representing reality for a while now. A strain's realistic potency ceiling under good cultivation is usually lower than whatever number moved the most product on a shelf -- and that's true even setting aside deliberate fraud, purely as a function of how labs have historically calibrated and rounded.

Worth a caution here too: alongside legitimate peer-reviewed findings like the Colorado study, there's a lot of viral-stat noise circulating in cannabis media and social feeds -- claims like 48% of products deviating by 20% or more that get repeated constantly but don't trace back to any identifiable study or dataset. Treat those the same skepticism you'd apply to an inflated label. If a number can't be traced to a named lab, agency, or published study, it's not evidence, it's just a number that sounds plausible.

The realistic takeaway for growers isn't that potency testing is worthless or that strong genetics don't exist -- well-bred cultivars absolutely can reliably hit strong, consistent potency under good conditions. It's that the useful signal isn't a single headline percentage on one bag. It's the trend across your own harvests, your own observed effects, and how a genetic line performs for you specifically, season over season, under your own lights and nutrient program. That data is more honest than any one lab's certificate, because it's not competing for your next testing contract.

How to Read a Certificate of Analysis Like a Skeptic

Photo by Charly Álvarez via Unsplash.

When a certificate of analysis lands in your hands, there are a few concrete checks worth running before you take the number at face value. First, confirm the lab holds ISO 17025 accreditation, the standard laboratory competence certification that at least establishes a baseline of documented procedure. Then check whether that lab has shown up in any recent state enforcement action -- suspensions, probation, or fraud allegations are public record in states like California, and a lab currently under scrutiny is not the one you want setting your expectations.

Next, know what an acceptable variance actually looks like to a regulator, not just to marketing. Massachusetts uses a permitted range of 75% to 125% of labeled potency as its enforcement benchmark -- meaning even the state acknowledges a wide legitimate band around any single number. If a product's label claims 25% THC, anything from about 18.75% to 31.25% would fall inside that official tolerance. That's a wider range than most buyers assume exists, and it's useful context before treating a single-point percentage as precise to the decimal.

If you're growing the same cultivar across multiple batches, compare potency results batch to batch rather than trusting any one report in isolation. Wild swings from the same producer growing the same genetics under similar conditions are a red flag -- either the producer is lab shopping for whichever facility reports highest that month, or there's real instability in the genetic line itself, which is its own problem worth knowing about.

It also helps to understand that total cannabinoid methodology alone can shift a number by several percentage points with zero fraud involved. Labs differ in whether they report decarboxylated THC directly or calculate total THC from a formula involving THCA and delta-9 THC, and that methodological choice alone produces different-looking numbers from chemically identical material.

For growers, the durable signal isn't any single lab's certificate -- it's consistent, well-documented genetics paired with your own trend data across harvests. That's part of why Seedtiva focuses on breeding and sourcing genetics with documented, repeatable performance: a predictable starting point matters more long-term than chasing whichever lab happened to report the highest number that week. Results will still vary with your climate, setup, and growing skill, but consistent genetics at least give you a stable baseline to measure against.

None of this variation is a fluke, and it isn't a minor rounding quirk that'll sort itself out. It's built into the current testing methodology at a structural level -- grinding, extraction, calibration, moisture correction, and methodology choices around decarboxylation all introduce error before anyone acts in bad faith. Layer commercial pressure to report the highest defensible number on top of that baseline noise, and you get a market where a 17-point spread on identical material isn't a scandal, it's just Tuesday.

What's changing is oversight, not the underlying chemistry. Massachusetts, California, and Maryland are building auditing infrastructure -- standardized procedures, independent re-testing, cross-lab benchmarking against reference facilities -- that arguably should have existed since state markets opened. Right now that enforcement is still happening lab by lab, case by case, which is why a BelCosta or a Verity Analytics makes headlines while plenty of other labs with similar practices haven't been caught yet.

Until that infrastructure matures nationally, the practical move for anyone growing or buying is simple: treat any single potency number as an estimate with real error bars attached, not a certified fact. Build your judgment on consistent, well-documented genetics and your own results across multiple harvests, rather than the highest percentage printed on a shelf. The plant's actual performance -- effect, yield, how it grows under your specific conditions -- will tell you more than any one lab's certificate ever will.