The 635 mg ceiling
This page does not claim to know what is inside a CogniMary capsule. It works out the most that can be, using the label's own printed total and a labelling rule the seller had to follow.
CogniMary's WMG-NVT Exclusive Blend declares 635 mg shared by eleven ingredients. Because 21 CFR 101.36(c) requires blend ingredients to be listed in descending order of weight, the ingredient in position k cannot weigh more than 635 ÷ k milligrams.
Applying that to the printed order, six of the eleven cannot reach the dose used in the research behind them. Berberine's ceiling is 105.8 mg against a 1,500 mg trial schedule. Two — lutein and zeaxanthin — fit comfortably.
The rule that makes this possible
A proprietary blend is a legal construction, not a marketing flourish. The regulation that governs it, 21 CFR 101.36(c), does two things at once, and the second undoes a little of the first.
What it permits
The individual amounts inside a blend do not have to be disclosed. The blend is identified by a name and the total weight of everything in it is placed on the same line to the right of that name.
635 mg, and no more detail
What it requires
The ingredients inside the blend must still be listed by name, indented beneath it, in descending order of predominance by weight. That ordering is not optional.
Heaviest first, lightest last
The ordering requirement is what turns a hidden split into a bounded one. If eleven weights run from heaviest to lightest and total 635 mg, then for any position k, the first k ingredients are each at least as heavy as the kth. So k times the kth weight cannot exceed the running total, which cannot exceed 635.
The whole derivation, in one line
If w1 ≥ w2 ≥ … ≥ w11 and they sum to 635 mg, then k × wk ≤ w1 + … + wk ≤ 635, and therefore wk ≤ 635 ÷ k.
That is it. No assumption about the formula, no estimate, no insider knowledge. An upper bound that follows from the label and the regulation.
The chart
The bar is the dose the ingredient was actually studied at. The amber tick is the most it can weigh in this blend at its printed position.
Federal labelling rules let a proprietary blend publish one total and hide the split. The same rules require the ingredients to be listed in descending order of weight. That second rule is the crack in the first one: if eleven weights fall from heaviest to lightest and add up to 635 mg, the ingredient in position k cannot weigh more than 635 ÷ k milligrams. Those ceilings are below, with the lowest dose each ingredient was actually studied at — the reading most favourable to the product.
Six of the eleven cannot reach a researched dose, and the arithmetic does not depend on trusting anyone. Berberine sits sixth, so it cannot exceed 105.8 mg; its usual trial schedule is 1,500 mg a day. Alpha lipoic acid sits eighth, ceiling 79.4 mg, against a 600 mg trial dose. GPC choline sits ninth, ceiling 70.6 mg, against 600 mg.
Two of them fit easily: lutein and zeaxanthin were studied at 10 mg and 2 mg. Ginkgo, in second place, has room for its 120–240 mg. Green tea has no cognition dose we could verify, so it gets no verdict.
Add up the lowest researched dose of every ingredient that has one and you get 4,492 mg. The blend is 635 mg. It would need to be about seven times larger — and this page still does not claim to know the split, because nobody outside the factory does.
Drag the slider and the whole chart moves. It is worth doing once: even at three times the printed weight, several ingredients still cannot reach the dose behind them, because the ceiling falls as fast as the position number rises.
Position by position
| # | Ingredient | Ceiling | Studied at | Verdict |
|---|---|---|---|---|
| 1 | Cinnamon Extract | 635.0 mg | 1,000 mg a day | Cannot reach |
| 2 | Ginkgo Biloba Extract | 317.5 mg | 120 to 240 mg a day | Can fit |
| 3 | Bilberry Extract | 211.7 mg | 160 to 480 mg a day of standardised extract | Low end only |
| 4 | Bacopa Monnieri Extract | 158.8 mg | 300 mg a day for at least twelve weeks | Cannot reach |
| 5 | Green Tea Extract | 127.0 mg | no cognition dose we could verify | No verdict |
| 6 | Berberine HCl | 105.8 mg | 500 mg three times a day, so 1,500 mg a day | Cannot reach |
| 7 | L-Theanine | 90.7 mg | 200 mg in a single dose | Cannot reach |
| 8 | Alpha Lipoic Acid | 79.4 mg | 600 mg a day | Cannot reach |
| 9 | GPC Choline | 70.6 mg | 600 mg a day | Cannot reach |
| 10 | Lutein | 63.5 mg | 10 mg a day in AREDS2 | Can fit |
| 11 | Zeaxanthin | 57.7 mg | 2 mg a day in AREDS2 | Can fit |
Scroll this table sideways →
Two rows need a word of explanation. Bilberry sits third, so its ceiling is 211.7 mg; the extract has been used at 160 to 480 mg, so the bottom of that range fits and the top does not. Green tea extract sits fifth with a 127 mg ceiling, but we could not verify a cognition dose for EGCG from a primary source, so it gets no verdict rather than a guessed one.
The six that cannot get there
These are the rows where the ceiling is below the researched dose, in the order they appear on the label.
Cinnamon extract — ceiling 635 mg
Heaviest position, so its ceiling is the whole blend. The trial we read used 1,000 mg a day for twelve weeks, and even if cinnamon were the only thing in the blend it would fall short. It is also studied for glucose and lipids rather than memory.
#1 of 11
Bacopa monnieri — ceiling 158.8 mg
The meta-analysis pooled trials at 300 mg a day, ran them for twelve weeks or more, and found an effect on speed of attention. This blend cannot reach half that dose at bacopa's printed position.
#4 of 11
Berberine HCl — ceiling 105.8 mg
Trials dose it at 500 mg three times a day. The ceiling here is a fourteenth of that. It is also a metabolic ingredient with, as far as we could establish, no human cognition evidence at all.
#6 of 11
L-theanine — ceiling 90.7 mg
The attention work uses a single 200 mg dose. Under half of that is available at position seven, and theanine is one of the few ingredients here with a clean acute effect to lose.
#7 of 11
Alpha lipoic acid — ceiling 79.4 mg
600 mg is the figure in the cognition pilot we read, and in the wider literature. The ceiling is an eighth of it. Cochrane found no qualifying randomised trials of ALA for dementia at all.
#8 of 11
GPC choline — ceiling 70.6 mg
The most encouraging single trial on this label used 600 mg a day for twelve weeks in people with amnestic mild cognitive impairment. Position nine leaves room for roughly a ninth of it.
#9 of 11
What this does and does not prove
It does not prove any of these ingredients is present at a low dose. It proves that none of them can be present at the dose the research used, because the blend is not large enough at those positions to hold it.
It also does not prove the product does nothing. It proves that if it does something, the reason is not that it reproduces those trials.
The arithmetic from the other end
There is a second way to look at the same number, and it is shorter. Take the lowest researched dose of every ingredient in the blend that has one, and add them together.
| Ingredient | Lowest researched dose |
|---|---|
| Berberine HCl | 1,500 mg |
| Cinnamon extract | 1,000 mg |
| Alpha lipoic acid | 600 mg |
| GPC choline | 600 mg |
| Bacopa monnieri | 300 mg |
| L-theanine | 200 mg |
| Bilberry extract | 160 mg |
| Ginkgo biloba | 120 mg |
| Lutein | 10 mg |
| Zeaxanthin | 2 mg |
| Total | 4,492 mg |
| The blend | 635 mg |
| Ratio | ≈ 7× |
Green tea extract is not in that table, because we could not verify a cognition dose for it, so its absence makes the total smaller and the conclusion more conservative rather than less.
Seven times. To give every ingredient the smallest amount any published trial used, the blend would have to be roughly 4.5 grams, which is not something two capsules can carry alongside 250 mg of magnesium oxide and a gelatin shell.
What a fair reading of this looks like
It would be easy to overclaim from this page, so here are the limits of what it shows, written by the people who made it.
We do not know the split
Nobody outside the factory does. Every number on this page is a bound, and a bound is not a measurement.
A ceiling, not a content
Some ingredients may not need the trial dose
Standardised extracts vary in strength, and a more concentrated extract can in principle do more per milligram. The trials we cite used specific materials at specific doses, and that is what they support.
A real caveat, stated
Formulas are not always additive
It is possible for a combination to do something its parts do not. It is also the oldest unfalsifiable claim in this industry, and no trial of this combination exists.
Possible, untested
What survives all three caveats is the original point: a buyer holding this bottle cannot check whether any ingredient is at a researched dose, and the ceiling says several of them are not.
What we would need to remove every doubt
- The split. Eleven numbers instead of one. That single change would make this entire page unnecessary.
- The extract ratios. A 10:1 ginkgo extract and a 50:1 ginkgo extract are not the same ingredient at the same weight.
- A certificate of analysis for this lot. Not a reference number — an actual document, from a laboratory, for lot COG-26/GO-2760.
- An identity test on the bilberry. Bilberry and blueberry are distinguishable in a laboratory. The panel's own naming leaves the question open.
None of those is exotic, and the first one is free. Until they exist, this page is the most anyone outside the company can say.
See what CogniMary costs
If you have read this page and still want to try it, that is a decision made with the arithmetic in front of you, which is the whole point of publishing it. Start with one bottle.
No price was published to this website. The figure is on the checkout page.
References
The sources behind the numbers on this page.
- U.S. Government Publishing Office21 CFR 101.36 — Nutrition labeling of dietary supplements — eCFR.gov · source
- Kongkeaw, Dilokthornsakul, Thanarangsarit et al. — 2014Meta-analysis of randomized controlled trials on cognitive effects of Bacopa monnieri extract — Journal of Ethnopharmacology · PMID 24252493 · source
- Snitz, O'Meara, Carlson et al. — 2009Ginkgo biloba for Preventing Cognitive Decline in Older Adults: A Randomized Trial — JAMA · PMID 20040554 · source
- Birks & Grimley Evans — 2009Ginkgo biloba for cognitive impairment and dementia — Cochrane Database of Systematic Reviews · CD003120.pub3 · source
- Higashiyama, Htay, Ozeki et al. — 2011Effects of l-theanine on attention and reaction time response — Journal of Functional Foods · DOI 10.1016/j.jff.2011.03.009 · source
- Klugman, Sauer, Tabet & Howard — 2004Alpha lipoic acid for dementia — Cochrane Database of Systematic Reviews · CD004244.pub2 · source
- Fiorentini, Cantatore, Ferrari et al. — 2023Cognitive and Mood Effect of Alpha-Lipoic Acid Supplementation in a Nonclinical Elder Sample: An Open-Label Pilot Study — International Journal of Environmental Research and Public Health · PMID 36767724 · source
- Lee, Kim, Park et al. — 2024Efficacy and safety of choline alphoscerate for amnestic mild cognitive impairment: a randomized double-blind placebo-controlled trial — BMC Geriatrics · PMID 39300341 · source
- Alzheimer's Drug Discovery FoundationBerberine — Cognitive Vitality For Researchers — alzdiscovery.org · source
- Ranasinghe et al. — 2025Effects of Cinnamomum zeylanicum extract on lipid profile, glucose levels and its safety in adults: a randomized, double-blind, controlled trial — PLOS ONE · DOI 10.1371/journal.pone.0317904 · source
- Chu, Cheung, Lau & BenzieBilberry (Vaccinium myrtillus L.), in Herbal Medicine: Biomolecular and Clinical Aspects — NCBI Bookshelf · NBK92770 · source
- Chew, Clemons, Agrón et al. — 2015Effect of Omega-3 Fatty Acids, Lutein/Zeaxanthin, or Other Nutrient Supplementation on Cognitive Function: The AREDS2 Randomized Clinical Trial — JAMA · PMID 26305649 · source
- European Food Safety Authority — 2018Scientific opinion on the safety of green tea catechins — EFSA Journal · DOI 10.2903/j.efsa.2018.5239 · source
The panel this page is built on
Everything on this page comes from two things visible in that image: the number 635 mg on the blend line, and the order of the eleven names underneath it.
No assumption has been made about the formula. No insider information was used, because none exists outside the manufacturer. The ceilings are what arithmetic allows given a printed total and a legally required ordering.
If the manufacturer publishes the split tomorrow, every number on this page becomes unnecessary, and we will replace it with the real figures and say so.
Why we built this instead of a star rating
Most product pages in this category end in a score out of ten arrived at by nobody in particular. This page ends in a division.
A score is an opinion
It compresses judgement into a number and hides the reasoning. Two people can read the same label and produce 9/10 and 4/10, and neither can be checked.
Not checkable
A ceiling is arithmetic
635 divided by 6 is 105.8 whoever does it. You can repeat it on a phone in ten seconds and get the same answer we did.
Checkable
A score ages badly
If the formula changes, the score is silently wrong. If the panel changes, the ceilings change with it and the working is visible.
Self-correcting
A ceiling has limits you can see
It is an upper bound. It says nothing about extract strength or combination effects, and this page says so rather than hiding behind a number.
Honest about itself
There is still a score on this site, on the home page, because readers ask for one. Every row of it is argued, and it is downstream of this page rather than the other way round.
Common objections, answered
- “The extracts could be concentrated”
- They could. A 50:1 extract at 60 mg is not a powder at 60 mg. But the trials we cite used specific materials at specific doses, and the label does not publish extract ratios for any of the eleven, so this objection replaces one unknown with another.
- “The combination works differently”
- Possibly. No trial of this combination exists, and none could be run without the split. Synergy is a reason to test a formula, not a reason to skip testing it.
- “The ordering rule might not be followed”
- If the ordering were wrong, the label would be non-compliant with 21 CFR 101.36(c). We have assumed compliance, which is the assumption most favourable to the product.
- “You cannot know it is not at the top of each ceiling”
- Correct, and we never say otherwise. The ceilings are upper bounds. Six ingredients fail even at their ceiling, which is the strongest version of the claim available.
- “Everyone in this category uses blends”
- Largely true, and it is an argument about the category rather than a defence of this label. The method on this page works on all of them.
We would rather publish the objections than wait to be told them. If there is one we have missed, [email protected].