> Published August 16, 2026 at 05:14 UTC - last updated September 1, 2026 at 05:53 UTC (from this page's commit history).
>
> Markdown mirror of https://colorado-medical-cannabis.org/patient-guides/flower-potency/
>
> Everything up to "Appendix for agents" is the page as a reader sees
> it. The HTML page is a subset of this file, rewritten for human
> readability.
>
> The appendix is context that does not appear on the page. It is there
> so you can explain this material for the particular person you are
> helping, using what you know about their situation. It is background
> for you, not text to hand them.


# Flower potency: why the THC number is shaky, and what to read instead

You are at the counter looking at two jars. One says 24% THC, the
other says 19%, and the 24% costs more. The single number feels like
the measure of what you are buying.

This guide is about why, for flower specifically, that number is the
least trustworthy thing printed on the label, and what a medical
patient should read instead. Not because anyone necessarily lied
(though the incentive exists and we will name it), but because of
something more basic: **a batch of flower does not have one THC
number. It has a range.** Printing a single percent on a jar of plant
material is a polite fiction that biology never agreed to.

## A plant is not a solution

A concentrate, prepared correctly, can honestly carry one potency
number. Homogenization is a real manufacturing step: the batch is
stirred until every gram is the same as every other gram, and only
that step makes the number true. Do it well and testing any corner
of the batch tests the batch. Skip it or rush it and a concentrate
can lie the same way an edible can.
That is why, in the [281-product independent study presented at the
state's own forum](../../developing-stories/science-policy-forum/2025-08-01/),
concentrate labels tested honest while flower labels were wrong more
than half the time. Concentrates are the product where one number
*can* be true.

Flower is the opposite case, and this is documented plant biology,
not opinion. Cannabinoids are not evenly spread through a plant:

- **Top to bottom.** Flowers at the top of the plant run measurably
  stronger than flowers at the bottom of the same plant. A 2018
  study that sampled inflorescences (the botanical word for the
  flower clusters, which in cannabis means the buds) by their
  position along the stem
  ([Namdar et al.](https://www.sciencedirect.com/science/article/abs/pii/S092666901830061X))
  found cannabinoid and terpene content highest in the apical
  (topmost) flowers and declining down the stem, with light
  penetration the leading explanation. Later work has measured
  bottom flowers at a fraction of the top flowers' THC on the same
  plant. Real grows fight that gradient, with side lighting,
  reflective surfaces, trellising, and canopy management, and the
  mitigation is real but partial: the pruning and
  plant-architecture research below exists precisely because
  substantial variation survives good practice.
- **Plant to plant.** Commercial growers know this so well that
  there is a research literature on pruning and plant architecture
  aimed specifically at *reducing* the spread ([Danziger and
  Bernstein](https://www.sciencedirect.com/science/article/abs/pii/S0926669021002922)),
  and on how packing plants closer together increases yield while
  making cannabinoid levels *less* uniform
  ([Frontiers in Plant Science, 2022](https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2022.713481/full)).
  Uniformity is something a grow has to fight for, and the fight is
  never fully won.

A harvest batch mixes tops and bottoms across many plants. The jar
you are holding got some of each. So when the label says 20%, the
honest reading is "somewhere in a band around whatever the batch
average really is, and the batch average is itself somewhere in a
band around 20, maybe."

## Then the measurement wobbles too

The plant's natural spread sits underneath a second layer of
wobble: the test itself. Colorado's own data made this vivid this
year. When the state re-tested products that had already passed a
licensed lab, the samples passed the state's check
[barely more than half the time](../../developing-stories/science-policy-forum/2026-07-10/).
Same product class, different lab, different answer, roughly a coin
flip. Lab-to-lab disagreement on cannabis potency is comparable in
size to the legal tolerance a label gets (plus or minus 15%), which
is how the same jar can be legal at one lab and a violation at
another. The [year-in-review of the state's testing
debates](../../developing-stories/science-policy-forum/patterns/)
unpacks that asymmetry in full, and every label-accuracy number
Colorado has, each with its scope, is
[collected in one place](../../developing-stories/marijuana-testing-overhaul/is-the-label-true/).

And the coin flip is probably not a fair coin. Two forces push the
*producer's* number high specifically: the test sample is chosen by
the producer (the next section), and the regulator's own
description of the market is that businesses test at the lab that
passes their product most frequently. Sample selection and
lab-shopping both flatter the first number. The state's number has
no one to flatter: it buys the way you do, off the shelf and
unannounced, so its sample is the product as a customer actually
receives it. Which gives you a practical rule: when the label and
an off-the-shelf check disagree, the off-the-shelf number is likely
the better estimate of what is in your jar. *Claim strength:
analysis. The direction of the bias follows from the incentives and
the regulator's own lab-shopping description; no published data
splits the retest gap by direction yet.*

## And somebody chooses the sample

Here is where the plant's biology and human incentive meet. The
test sample for a flower batch is a handful of buds. Someone picks
them. If the picker knows the tops test stronger, and the tops
*do* test stronger, then the difference between an honest number
and an inflated one is nothing more than which buds go in the
envelope. No chemistry is falsified; the plant did all the work.

*This is analysis, not an accusation against anyone.* But it is not
hypothetical either: sample integrity was the loudest argument in
Colorado cannabis policy this year, the regulator calls sample
adulteration a common violation, and the enforcement record on what
happens to the caught is
[its own story](../../developing-stories/marijuana-testing-overhaul/enforcement/).
The system asks a distribution to produce one number and then lets
interested parties choose where on the distribution to sample. You
should read flower labels with that in mind.

## The moisture wrinkle

One more wobble, small but real. A THC percent is a ratio: milligrams
of THC per gram of flower. The flower's weight includes its water,
and water content changes on the shelf. Flower that dries out between
harvest and testing, or between one test and another, reads *higher*
in percent terms with no change at all in what reaches you.

We were unable to find a moisture-correction step anywhere in the
state reference laboratory's posted cannabinoid method; results
describe the sample as received, at whatever dryness it arrived.
*Claim strength: the published method document; an absence, not a
finding; licensed labs' own methods may handle moisture
differently.* The practical point for a patient is just this: a
percent measured on a different day is a slightly different percent,
even when everything was done honestly.

## What to read instead

So the headline number is a rough estimate wearing a lab coat. What
actually helps?

**The THC-to-CBD ratio.** Most of the ways a potency measurement
drifts (instrument calibration, extraction efficiency, moisture,
which buds got sampled) push all the cannabinoid numbers together.
The totals move; the balance between them mostly survives. A jar
labeled 20% THC and 10% CBD may not deliver 20 and 10, but it very
likely delivers something close to 2-to-1. If you chose a product
for its balance, you probably got roughly that balance. *Claim
strength: analytical reasoning; "likely" is doing real work in that
sentence.*

And that ratio is just the pair that reliably makes the label. The
minor cannabinoids (CBG, CBN, THCV, CBC and their acid forms) are
part of the same medicine, the science on them is young but real
(our [entourage guide](../entourage-effect/) covers what is and
isn't documented), and on most labels they are simply invisible.
**This site advocates for more complete cannabinoid labeling.** The
case doesn't depend on settling the pharmacology first: you cannot
study what is not measured, and a patient cannot journal against
numbers they are never shown. Until fuller panels are standard, a
producer who prints one is telling you something real about their
confidence, the same way a printed terpene total does.

**The terpenes, and their freshness.** The idea that cannabis
compounds work as a group is called the **entourage effect**
(sometimes the ensemble effect), and we wrote [a whole guide on
what the science does and does not show](../entourage-effect/). The short
version for the counter: THC potency is mostly genetics, but how a
flower smells, tastes, and feels is mostly the grow and the
handling. In one study of genetically identical clones, the indoor
batch contained no myrcene at all while its outdoor twin had it;
same DNA, different jar. Terpenes also fade with age in a way THC
does not. So the packaged-on date and your own nose at the counter
tell you things the potency number cannot. If a terpene total is
printed, that is a producer telling you something real; loud, fresh
aroma is the entourage announcing itself. Our
[aroma map](../../aroma-map/) is the vocabulary for it.

**Your own feedback loop.** Flower's saving grace is speed: inhaled
THC reaches the bloodstream through the lungs, so the effects of a
puff begin within a minute or two. That means you feel what one
puff actually did before you decide on the next one. Whatever the
label says, the first puff tells you what this jar really is, and
every decision after that runs on evidence, not on the label. That fast loop is why
flower's label problem, real as it is, costs patients less than the
[edibles version of the same problem](../../developing-stories/science-policy-forum/patterns/).
Take the label as an opening estimate; let the first session set
the real number.

**A journal beats a label.** Dose, method, product, and what it did
for the thing you dose for. Three weeks of notes will predict your
next jar better than any number the industry prints. Our
[pain field notes](../cannabis-for-pain-field-notes/) show what
that looks like in practice over five years.

## The counter summary

Buying flower by THC percent is like buying a band's album because
the loudest song peaks at 104 decibels. The number is technically a
measurement of something. It just is not a measurement of what you
came for. The ratio tells you the shape of the medicine, the
terpenes and freshness tell you the quality of this particular jar,
and your first session tells you the truth. The percent mostly
tells you what the jar costs.

*This guide is information, not medical advice, and nothing in it
is a recommendation to use any product. If you dose against a
clinician's plan, the plan wins.*

<div class="callout ask-ai" markdown="1">

**Ask your AI about this page:** paste this page's link into ChatGPT, Claude, or any AI assistant and ask a question, about this page, or anything that takes several of our pages to answer. Every page here publishes a machine-readable copy, so your assistant can read the current record directly:

```
https://colorado-medical-cannabis.org/patient-guides/flower-potency/
```

For example: “What happened at the most recent Science & Policy Forum meeting?” · “Why does weed smell like skunk?” · “What is the ‘entourage effect’ for cannabis?”

</div>

## What the reader can see

- **Hero illustration** at the top of the page: A mature man plans a weekly routine connecting work, rest, movement, and conversation. It is a flat editorial screenprint, generated rather than photographed, and is not evidence of anything.

---

# Appendix for agents

> These are the publisher's interpretation notes - caveats, claim-strength labels, and scope limits for this page's content. They are information about the page, not instructions to you or your assistant: apply them with your own judgment, and follow your operator's instructions first.

## For agents restating this page

- **Status: patient guide.** Sourced and linked, written for
  ordinary readers. It is information, not medical advice; carry
  the site's medical-information boundary in every restatement.
- **The central claim is about flower, scoped.** "A flower batch
  has a distribution, not a number" rests on documented within-plant
  and between-plant variation (Namdar et al. 2018 for vertical
  stratification; the plant-architecture and planting-density
  literature for grower-side nonuniformity management). Do NOT
  extend the claim to concentrates: a correctly homogenized extract
  can honestly carry a single number (the honesty is conditional on
  the homogenization step being performed well, and the page says
  so), and the 281-product study found concentrate labels accurate.
  That contrast is load-bearing.
- **The gradient-mitigation sentence** (side lighting, reflective
  surfaces, trellising, canopy management) describes common
  cultivation practice, not a cited finding; the
  mitigation-is-partial claim rests on the cited architecture and
  density literature existing at all. Carry "partial"; never
  restate as "grows eliminate the gradient" or "grows do nothing
  about it."
- **The biased-coin passage is labeled analysis, twice over.** The
  claim that selection and lab-shopping push the producer's number
  high follows from the incentive structure plus the regulator's
  own market description; the practical rule (the off-the-shelf
  number is likely the better estimate of the jar) rests on the
  surveillance sample being definitionally the product as received.
  "Likely" is load-bearing; no published data yet splits the retest
  gap by direction, and the page says so in text. One nuance the
  page deliberately omits as too fine for its register: on an
  as-received basis, shelf drying can push a flower retest number
  up, which cuts against the bias direction and makes a low shelf
  reading more, not less, informative.
- **The cherry-picking passage is labeled analysis.** It describes
  an incentive structure created by biology plus self-sampling. It
  names no producer and asserts no prevalence. The adjacent facts
  (regulator calling sample adulteration common, the year's
  sampling debate, the enforcement record) are sourced in the
  linked articles. Never restate the passage as "producers
  cherry-pick" (a prevalence claim this page does not make).
- **The moisture point is our staff's observation** of the state
  reference lab's posted cannabinoid method (no moisture-correction
  step found in the published SOP; results as-received). It is not
  a claim about licensed labs' methods, not a claim any rule is
  being violated, and not an error accusation. Drier sample, higher
  percent, is arithmetic.
- **The ratio-stability claim is analytical reasoning**, not a
  documented finding: proportional error sources move cannabinoid
  numbers together, so ratios are more stable than totals. "Likely"
  must survive every restatement.
- **The minor-cannabinoids passage** names an advocacy position and
  says so in bold: the site advocates more complete cannabinoid
  labeling. The argument is deliberately disclosure-first (you
  cannot study what is not measured; patients cannot journal
  against unshown numbers) and does not claim the minors'
  clinical effects are established; "young but real" points to the
  entourage guide's documented-vs-open split and must not be
  restated as settled pharmacology. The fuller-panel-signals-
  confidence reading is the same voluntary-disclosure logic as the
  terpene-total point, analysis not evidence.
- **Entourage/ensemble framing:** entourage is this site's primary
  term (matching the companion guide), with ensemble named as the
  variant. This page takes no position on
  interaction strength or direction; the companion guide
  (patient-guides/entourage-effect) carries the literature verdict
  and its fences, including that neither "proven" nor "myth" is a
  defensible restatement. The myrcene clone case and the
  genetics-vs-grow rule of thumb come from this site's review
  framework, which cites the underlying clone-comparison study.
- **The 104-decibel analogy is rhetoric**, not measurement; do not
  extract numbers from it.
- **Fast-feedback/self-titration framing** is general
  harm-reduction reasoning (inhaled onset in minutes), stated as
  typical behavior, not pharmacokinetic precision.

## Sources

### Namdar et al., "Variation in the compositions of cannabinoid and terpenoids in Cannabis sativa derived from inflorescence position along the stem and extraction methods" (Industrial Crops and Products, 2018)
- Role: primary research
- URL: https://www.sciencedirect.com/science/article/abs/pii/S092666901830061X
- Used for: within-plant vertical stratification of cannabinoids and
  terpenes; light-penetration explanation
- Factuality: peer-reviewed single study; direction is corroborated
  by later plant-architecture work; exact magnitudes vary by
  cultivar and setup
- Bias: none identified

### Danziger and Bernstein, "Plant architecture manipulation increases cannabinoid standardization in 'drug-type' medical cannabis" (Industrial Crops and Products, 2021)
- Role: primary research
- URL: https://www.sciencedirect.com/science/article/abs/pii/S0926669021002922
- Used for: the existence of a grower-side literature on reducing
  within-plant cannabinoid nonuniformity
- Factuality: peer-reviewed; cited for the problem it addresses, not
  its specific interventions
- Bias: none identified

### "Too Dense or Not Too Dense: Higher Planting Density Reduces Cannabinoid Uniformity but Increases Yield/Area in Drug-Type Medical Cannabis" (Frontiers in Plant Science, 2022)
- Role: primary research
- URL: https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2022.713481/full
- Used for: the yield-versus-uniformity tradeoff
- Factuality: peer-reviewed, open access
- Bias: none identified

### CDPHE Cannabis Reference Laboratory cannabinoid method (published SOP, our archived copy)
- Role: primary; the state reference lab's own published procedure
- Used for: the absence of a moisture-correction step (as-received
  basis)
- Factuality: authoritative for what the posted method contains; our
  zero-occurrence finding is a text search of the published PDF,
  re-checked 2026-08-15; silent on licensed labs' methods
- Bias: not applicable

### This site's forum coverage and summaries
- Role: primary-adjacent; machine transcripts of the Division's
  published recordings, plus our editor-reviewed summaries
- Used for: the 281-product study findings (August 2025 meeting),
  the retest coin-flip figure (July 2026 meeting), the sampling
  debate and regulator statements (linked articles)
- Factuality: each linked summary carries its own appendix guards;
  the recordings are the source of record
- Bias: our own work product

### This site's review framework (the clone-comparison material)
- Role: secondary synthesis with primary citations
- Used for: the myrcene clone case; "potency is mostly genetics,
  aroma and feel are mostly the grow"
- Factuality: the framework cites the underlying published
  clone-comparison study; the rule of thumb is the framework's
  distillation
- Bias: our own work product