What a Cannabis COA Cannot Tell You About Your Products

Nevada’s Cannabis Compliance Board fined a lab called Cannex Nevada, doing business as LettuceTest, $62,500 and pursued a nearly ten-year ban from the industry. The lab was accused of passing microbially contaminated product and inflating potency numbers (Lange, 2022).

California’s cannabis regulator revoked or suspended licenses for at least four labs between December 2023 and mid-2024. One lab, ProForma Labs of Salinas, missed a banned pesticide called chlorfenapyr entirely. It also reported THC at 24% when the true level was 16%. Another lab, Verity Analytics, was accused of inflating results by as much as 32%. Its director reportedly cleared pesticide screens by visual inspection alone. California’s licensed lab count fell from 37 to 27 over the course of 2024, a drop of more than a quarter (Roberts, 2024).

In Massachusetts, an undercover audit released in August 2026 found that a fifth of tested products significantly overstated THC. Regulators cited 13 companies for products at least 25% weaker than their label claimed. Eight of those were off by more than 30%. Two companies were off by 45%, and one preroll was 55% weaker than advertised (Hecht, 2026).

How does a number get inflated? Sometimes the method is crude. THC is measured as a percentage of a sample’s dry weight. Over-drying a sample before testing pushes the number up. Most states do not require moisture to be reported alongside potency, which makes this trick invisible on the certificate.

Sometimes the problem is sample selection. Labs and cultivators can submit an unrepresentative, hand-picked bud rather than a random sample from the batch.

Increasingly, industry reporting describes outright lab shopping. That means submitting the same product to several labs and keeping only the most flattering result (Erickson, 2024). Massachusetts has since moved to require that every sample from a batch go to a single lab. This followed a lawsuit in which one testing company accused eight competitors of manipulating results (Robertson, 2025).

What the COA Panel Was Never Built to See

Even an honest cannabis COA has blind spots built into its design, not its execution.

Terpene Profiles

Terpenes are the clearest example. Testing for them is optional in Colorado and Michigan. California tests for terpenes only when a licensee requests it, not as a blanket rule for every product.

This means a COA showing “similar cannabinoids” between two products may say nothing about the compounds most responsible for flavor, aroma, and any entourage-style effect on the high.

Pesticides

Pesticide screening has a similar limit built in. California’s required pesticide list covers dozens of compounds, but anything outside that list does not appear on the report, whether it is present or not.

A 2024 investigation tested 42 legal California products for a broader panel than the state requires. 25 of them exceeded safe limits for at least one compound. Some vape products carried concentrations of banned pesticides many times higher than federal tobacco allowances, despite holding a passing COA (St. John & Halperin, 2024).

Toxicologist Max Leung of Arizona State University and independent lab chemist Josh Swider were both interviewed. Both said the testing system has few real checks once a product clears its narrow required list.

Remediation

Then there is remediation, the practice of treating contaminated cannabis, usually with radiation or ozone, to force a failing microbial test to pass instead of destroying the batch.

A 2026 investigation found the practice widespread in Maine. One industry estimate put roughly half the state’s recreational supply through some form of treatment. Maine has a labeling rule for remediated products on the books. State regulators confirmed to reporters that they have stopped enforcing it, after industry pushback that a label reading “irradiated” would scare off customers without telling them anything useful about safety (Tusinski, 2026).

Yasha Kahn, the MCR Labs co-founder whose commercial stake in that analysis was already flagged, told the same reporter that radiation and ozone treatment do not reliably keep mold from coming back later in a treated batch (Tusinski, 2026). His lab also profits from being seen as the more rigorous alternative, a stake worth weighing alongside his claim.

A passing retest only shows that mold was not detected that day. It does not prove the underlying contamination is gone for good. None of that history travels with the final product. A remediated batch’s COA looks identical to one from flower that was never contaminated at all.

Weed Storage

Storage creates a quieter version of the same problem. THC breaks down toward CBN over time. Terpenes evaporate with heat, light, and time. A COA reflects the plant at the moment it was tested, not the moment it is finally sold, sometimes weeks or months later off a dispensary shelf.

Environmental Footprint

Cultivation’s environmental footprint sits even further outside anything a COA measures.

A cannabis plant grown outdoors can use roughly six gallons of water a day across a five-month season, about twice what comparable field crops need. Indoor cultivation trades water for electricity. Lighting, climate control, and dehumidification account for the large majority of an indoor grow’s power use, and producing one kilogram of indoor flower has been estimated at roughly 4,600 kilograms of carbon dioxide (Zheng et al., 2021).

None of that shows up anywhere near a certificate of analysis, and nothing on the certificate speaks to whether a plant was grown indoors or in the sun at all.

Cannabis COAs Test One Sample, Not the Whole Batch

Even a perfectly honest, broad COA describes one sample pulled from a defined batch. It does not describe every unit sold under that batch number.

Flower is not uniform. Cannabinoid content and moisture vary depending on where a bud sat on the plant and how it was dried and handled afterward. Regulators in New York have built their sampling rules around this kind of variation, spreading samples across a batch rather than pulling from one spot (New York State Office of Cannabis Management, 2022).

A representative sample, honestly drawn, still cannot promise that the exact jar a patient buys matches the jar that got tested.

How to Fix Cannabis COAs for Consumers

The reform effort for cannabis COAs is real, if scattered.

The Cannabis Regulators Association is a coalition of state regulators formed to coordinate standards in the absence of federal guidance. Its leadership has pushed for independent state reference labs that can audit and spot-check commercial labs’ work. As of 2025, only 16 of 42 cannabis-regulating states had one or planned to build one (Rowińska & Bajak, 2025).

Two technical groups are working on this too. ASTM International’s Committee D37 and AOAC International are both developing voluntary standard methods for sampling and testing. Adoption remains state-by-state and optional.

California’s license revocations and Massachusetts’s undercover audit both work after the fact, catching bad actors once product is already on shelves. The reference-lab and voluntary-standards efforts aim earlier, at the testing process itself, but progress there has been slower than the enforcement actions targeting individual labs.

Investigative journalism has done more to expose these gaps than any single regulatory body so far. Leafly exposed the Washington case. MJBizDaily has kept covering lab shopping nationally. The Los Angeles Times and WeedWeek broke the California pesticide story. The Portland Press Herald uncovered Maine’s remediation practices. The Boston Globe ran the Massachusetts audit. Together, this reporting has pushed regulators toward action more than any single oversight program has, largely because no state yet runs one with real teeth.