Who this hits: Colorado's testing decisions read from six different chairs

A rule change is not one event. The same decision that barely touches a dispensary can reshape an extraction line's week, and the question a lab finds most alarming may be the one a patient most needs answered. This page takes the decisions and non-decisions from Colorado's testing overhaul (including the August 7 Science & Policy Forum) and reads them from six positions in the supply chain.

The underlying facts are sourced on the linked pages. What each decision means for each group is our analysis, labeled as such throughout, and kept separate from what the state actually said.

The decisions, in one place

Settled, at least for now:

1. Third-party sample collection will not come from the forum. The Science & Policy Forum reached consensus not to recommend requiring an independent party to pull compliance samples. Related sampling suggestions continue inside rulemaking, separately.
2. Method-validation rules stay as they are. The state's review found the criteria "comprehensive and sound"; only clarification is proposed.
3. Sample manifests now show the source package size for test-batch transfers to labs, with no system check that a large enough sample was pulled.
4. Surveillance-testing findings move into rulemaking, out of the forum.

Unsettled, on the record:

5. Why labs match each other but not the state's lab. Colorado labs agree closely in proficiency testing, then diverge from the state reference lab when products are pulled off shelves and retested. The state's answer: "I don't think we have an answer to that."
6. Whether tighter definitions fix anything, absent a single standard method every lab must use.
7. Rule or guidance: whether tightened expectations become regulation or stay in the audit checklist.
8. How far matrix-specific quality control should go, given how many forms concentrates and edibles take.
9. How to test partially infused edibles, where a procedure changed ahead of its rulemaking.
10. When the expanded conversion-solvent screen switches on: it is law, but no lab is certified to run it yet.

Medical cannabis patients

Analysis. Patients are the group with the most at stake in numbers being right, and the least ability to check them.

• Reproducibility matters more to you than potency. Someone managing pain, spasticity or appetite finds a dose that works and needs the same product to behave the same way next month. Decision 5 (the unexplained gap between what a lab reported and what the state's lab found) is precisely a reproducibility question, and it is open.
• The dose-evenness rule now speaks your language. Edibles must pass homogeneity testing so the dose is spread evenly through a batch, and the January 2026 revisions changed the threshold language from THC to cannabinoids, which finally matches the THC-with-CBD, CBN or CBG products many patients specifically seek out.
• Concentrates carry a distinct risk you should know about. The state warned in January that concentration processes concentrate pesticides too: plant material can pass its pesticide test and become a concentrate that fails. If your regimen relies on concentrated forms, this is the contamination pathway to understand.
• The person who made your medicine still pulls the sample that proves it is safe. Decision 1 leaves self-sampling in place. Nothing got worse; the change that would have addressed sample manipulation is not coming from this body.
• Practically: ask for the certificate of analysis, prefer stores that hand it over without friction, and watch the state's health-and-safety advisories, which the Division says have increased alongside pesticide cases.

Recreational consumers

Analysis. The buying decision is usually driven by the potency number, which is exactly the number under question.

• The label is a batch average, not your gram. Potency is reported as an average across samples from the whole batch. That is the rule working as designed, not a defect. But it is not a measurement of the jar in your hand.
• Decision 5 is your decision too. When the state pulls products off shelves and gets different numbers than the original lab, the reliability of the number you shop on is what is in question. Nobody claimed to understand the gap.
• The screen aimed at converted hemp cannabinoids is not running yet. The expanded solvent panel targets solvents used to convert CBD into intoxicating cannabinoids. It is law with no certified lab to perform it, so today it protects nobody, and the state will give 72 hours' notice before enforcing it.
• Surveillance testing is the closest thing you have to an advocate. It is the program that buys products like a customer would and retests them; its findings are now feeding rulemaking.

Cultivators (growers)

Analysis. Growers get the mildest version of the method fight and the sharpest version of the liability question.

• You keep collecting your own samples. Decision 1 means the cost, scheduling control and compliance responsibility of self-sampling all stay where they are. Related sampling changes in rulemaking are still live, so this is relief, not closure.
• Your sampling is now visible to the lab. Decision 3 puts the source package size on the manifest: a 100-gram package yielding a 2-gram test batch is now legible to whoever receives it. Nothing rejects an undersized sample automatically, but the number is no longer private.
• Test by category as well as by strain. The January revisions clarified that potency testing follows each category you cultivate, not only each strain.
• A pass on your flower is not a pass downstream. Pesticide residues concentrate during extraction, so material that passed at your grow can fail once it becomes concentrate. A failed pesticide test leaves the extractor two options only: retest the original samples twice, or destroy the batch. They cannot clean it up. Because that loss falls entirely on them, expect buyers to protect themselves before they purchase: asking to tour your operation, requiring test results and sampling plans up front, putting product-safety terms in the contract. No rule requires you to agree to any of that. Your customers will ask anyway. See the friction this creates.
• Flower is the easy matrix. In the debate over matrix-specific controls (decision 8), flower is the form everyone agrees is straightforward. The complexity (and any new burden) lands downstream.

Testing laboratories

Analysis. Labs are the group facing the most change and the most unwelcome attention.

• No overhaul of validation rules, but expect tightening. Decision 2 keeps the framework. The state named specific candidates: a common reference method for establishing detection limits, calibration expectations beyond the single R² value in rule, trend tracking over time, and clearer matrix-control definitions.
• Rule or checklist changes what non-compliance means. Decision 7 is not cosmetic: guidance in an audit checklist and a requirement in rule carry different consequences, and labs have a direct interest in which one this becomes.
• The divergence question is a reputational exposure. A lab director asked plainly why labs cluster in proficiency testing and diverge in surveillance testing, and said the results make labs nervous. The state has no answer yet. Until it does, the gap is available for anyone to interpret uncharitably.
• The solvent-panel delay is squarely at your feet. Implementation waits on certification, the Division has urged labs to prioritize it, and it has said it will "assess rule updates to prevent implementation delays to future testing requirements."
• Skepticism was voiced from inside the room. Without a standard method every lab runs, one participant argued, differences will persist no matter how the definitions are sharpened, and labs cannot see each other's methods to locate where divergence begins.

Extractors and product manufacturers

Analysis. This group carries the heaviest load in the overhaul and has the most exposure to decisions still pending.

• Pesticide failures have no cleanup path. Retest the original samples twice and pass, or destroy the production batch. No remediation, no dilution into a new batch. The Division wrote a bulletin because it observed exactly that being attempted.
• Sourcing is your real control. Since residues concentrate during extraction, contamination risk is decided when you choose plant material, not when you test. The state's recommendations are concrete: audit suppliers, require proof of passing results, review sampling protocols, inspect what arrives against the certificate, and put safety terms in contracts. Doing that well is a relationship problem as much as a compliance one. See the friction this creates.
• Your reduced-testing allowance is narrower than you may think. A concentrate made with different hardware (a different cartridge size, design or manufacturer) counts as a different product that must earn its own allowance. Changing your plant-material supplier can warrant re-authorization.
• Partially infused products are in genuine limbo. Decision 9: a procedure covering non-infused and partially infused edibles changed before the matching rulemaking, and labs are unsure how to test them. The state took the issue away to review. If you make these, this is the item to track.
• Matrix-specific QC could multiply your labs' work, and your bills. One estimate voiced in the meeting was roughly twenty control samples per batch if matrix categories were specified narrowly. Decision 8 is unresolved, and the state acknowledged the risk of writing rules too specifically.
• Mycotoxin testing closed a gap for some concentrates. Water-based and heat/pressure-based concentrates now require mycotoxin testing when the wet plant material was not tested for microbials.

Retail stores and dispensaries

Analysis. Retailers did not cause any of this and absorb most of its public consequences.

• You sell what other people tested. Recalls and health-and-safety advisories arrive at your shelves, and the customer conversation happens at your counter, not at the lab.
• Decision 5 is a customer-trust problem before it is a science problem. When potency confidence wobbles, the complaint lands on the budtender. Staff who can explain that a label is a batch average (and that the state is openly working an unresolved question) are worth more than staff who cannot.
• Hemp products on your shelf are inside the tent. Any hemp product entering the regulated market, as an ingredient or for sale, must be tested at a state-licensed marijuana testing facility.
• Certificates of analysis are becoming a merchandising asset. As testing questions surface publicly, the ability to produce the paperwork on request is a differentiator, particularly for medical patients, who have the strongest reason to ask.
• Medical and retail run on parallel licenses. Colorado licenses medical and retail testing separately, and the requirements track each other closely; a store operating both sides manages two versions of the same obligation.

The friction this design creates: growers and extractors

Analysis. Read the pieces together and a pattern shows up that no document states outright: a meaningful part of how grow-side quality gets policed is by making the concentrate maker police it.

The state does not have an inspector standing in every cultivation room. What it has is a rule putting the entire loss on the extractor when a grower's material fails after extraction, and guidance telling extractors to audit their suppliers, demand test results and sampling plans, inspect deliveries against the paperwork, and write product-safety terms into contracts. A third lever points the same way: changing which farm supplies you can require re-authorizing a reduced-testing allowance, so your sourcing decisions carry regulatory consequences too. The buyer becomes the inspector because the buyer is the one who pays when nobody inspects.

Then comes the sentence that decides how much this can hurt. In the same bulletin that recommends writing those contracts, the Division states plainly that it "does not engage in contract disputes between Licensees." Colorado law makes agreements between licensed cannabis businesses enforceable, so when one of these arrangements breaks, the venue is a courtroom, not the regulator. The state supplies the pressure and stays out of the argument.

How this hurts a grower. Refusing an audit costs sales to a competitor who accepts one. Accepting a broadly written contract can expose you to a claim far larger than the flower was worth: the extractor's loss includes their processing and finished product, not the price they paid you. And there is no appeal to a regulator, because the regulator has said this isn't its fight.

How this hurts an extractor. An indemnity from a thinly capitalized grower is worth very little at the moment you try to collect; you ate the destroyed batch either way. Police too hard in a tight supply market and you lose access to material. And you have no authority to compel anything: your only real sanction is refusing to buy.

The trap that catches both. Blend material from several farms into one production batch and a pesticide failure destroys all of it while proving nothing about whose material caused it. Single-source production runs keep attribution possible; blended runs make every supplier a suspect and none of them provably responsible. That is a decision made before the run, not after the result.

An asymmetry worth naming. After a failed pesticide test the choice (retest the original samples twice, or destroy) belongs to the extractor alone. The grower has no standing in a decision whose outcome determines whether they get blamed.

The structural consequence. A company that owns its own farms does not have this problem at all. When the cost of a contaminated input lands entirely on the buyer, and the regulator declines to arbitrate, bringing cultivation in-house is a rational hedge rather than an ambition. That is worth watching as a possible quiet contributor to independent grows being absorbed into larger vertically integrated operations, a pattern visible in Colorado without anyone having designed it as policy.

What managing it well looks like. Settle it before the first delivery, not after the first failure: who tests what and when; whether incoming material is checked before processing, who pays for that, and how it is structured (the Division has separately warned that research-and- development testing must not be used as a substitute for required compliance testing); a liability cap proportional to the sale rather than to finished-goods value; proof of insurance; an agreed dispute process such as arbitration, given that the state will not referee; and single-source runs wherever attribution matters. The parties who write this down in advance will do fine. The ones who discover it during a destroyed batch may not.

What to watch next

The forum meets again on September 4, 2026, with sample collection procedures on the agenda, the topic several August 7 questions were deferred into, and the one closest to the sampling question that just came off the table. A separate technical session on the Aspergillus method is being scheduled. And the conversion-solvent screen switches on when the first lab is certified, with 72 hours' notice.

If your position in this chain isn't represented above, or we have read your part of it wrong, tell us: editor@colorado-medical-cannabis.org. Corrections from people who do the work are the best kind.