The LabelTruth Methodology

Mathematical transparency in supplements — how we score, and why.

Medical disclaimer. LabelTruth is an independent data-analytics platform, not a medical advisor. We calculate mathematical coefficients (PPAA) from publicly available scientific literature and EFSA standards. The figures shown are objective pharmacokinetic derivations, not treatment recommendations. We do not provide medical diagnoses or treatment plans. For any health concern, deficiency, or condition, always consult a qualified healthcare professional.

1. The problem with supplement labels

The supplement industry runs on information asymmetry. A label may read “400 mg of Magnesium”, but depending on the chemical form your body may absorb very different amounts of it. The label tells you how much compound is in the jar. It does not tell you how much reaches your bloodstream, how many units you need per day, or whether the form itself is still under review.

LabelTruth exists to close that gap. We strip the marketing away and evaluate each product on biochemical and logistical facts alone. Our methodology is public, and when a public comparison layer is not yet defensible, we remove it instead of inventing a softer replacement.

2. PPAA — Price Per Actually Absorbed

PPAA remains an internal comparison metric in the database. It is not currently shown as a public scorecard or as a product badge, because several older verdict thresholds were retired rather than defended. The underlying calculation still moves from label to bloodstream in four steps:

Elemental mass. Every supplement is a chemical compound (e.g. magnesium bisglycinate). We remove the carrier-molecule mass by molar fraction to obtain the pure elemental weight of the active ingredient per serving.
Bioavailability coefficient. We never assume 100% absorption. Drawing on in-vivo clinical studies indexed in PubMed, we apply a baseline fractional-absorption coefficient to each chemical form (for example, roughly 0.04 for oxides and around 0.25–0.35 for chelates).
PPAA. Dividing the retail price by the mass that actually reaches the bloodstream gives the price per 1 gram (1000 mg) of absorbed active ingredient. This exposes products that are cheap at the register but biologically expensive.
30-day cost. On the product page this is the cost of the labelled serving for 30 days. The Target line shows the cost of reaching the target — the amount we treat as the goal for what comes from a supplement, which is not the same number as the label reference value. Manufacturer serving suggestions are often understated (“fairy dusting”). Where no target has been decided for a form, the Target line states no 30-day figure.

Formula: PPAA = Retail Price ÷ (Total absorbed elemental mass in the container) × 1000. Prices are normalized to a single currency (EUR) before any comparison.

3. Public comparison policy

Some earlier public verdicts were built on anchor thresholds that were not strong enough to keep as user-facing facts. Those verdicts are being removed, not renamed. That includes public matrix-category labels, named bioavailability verdicts, public “save X% vs market median” copy, and cross-form value leaderboards that imply a settled ranking where the underlying assumptions are still moving.

What remains public today:

Chemical form identity. We show which form a product uses, or that the form is still unresolved.
Element unit price. Within one chemical form we rank by euro per 1000 mg of the element, not by an absorption-adjusted score. For elements dosed in micrograms we show price per 1000 µg instead of per 1000 mg. The calculation is identical; only the scale differs. Which elements those are is our own editorial choice, recorded per element in the catalogue — not a threshold computed from the data. For protein powders we show price per 100 g of declared protein. The calculation is the same; the unit is the label's own.
Units a day to the target. We show how many pills, scoops, drops, or gummies are needed to reach the target. Where a single serving already covers that target, we do not invent a fractional daily regimen. We state N× per serving instead: one serving provides N× the target.
30-day cost of the target. We show the cost of reaching the target for 30 days only when that takes at least one unit per day. Below that, the euro figure would price a regimen nobody follows, so it is withheld until we have the label's own daily serving.
Evidence status. Element hubs distinguish between reviewed and unreviewed form rows.

Important consequence: a product can lose a public label even when its underlying numeric inputs did not change. That is not a hidden downgrade. It is the removal of a public interpretation layer.

Units a day

Bioavailability is moot if a product cannot be taken every day. We compute how many units a day it would take to reach the target, and we show that number on the product card. We do not turn it into a label. A threshold saying “more than four is too many” would be a decision, not a measurement, and thresholds of that kind were removed from this page for the same reason.

The figure is also used internally, where it does not become a claim about the product: a serving instruction implying dozens of units a day is more likely a label we misread than a product anyone takes, and those rows are withheld from ranking rather than badged.

Why the target dose is not the label reference value

A Nutrient Reference Value is a labelling comparison figure under EU Regulation 1169/2011, and it describes total intake including food. An EFSA tolerable upper intake level is also total intake, unless the opinion limits it to supplements. Magnesium is that case: the reference value is 375 mg of total intake, while the upper level for supplemental magnesium is 250 mg. Using the reference value as a supplement goal therefore sets a target that exceeds the safe supplemental ceiling.

We now keep the two apart. The units a day and the 30-day cost of the target are computed from a target that may never exceed the upper level, and every target records what it is derived from — an upper level, an EFSA adequate intake or population reference intake, or, stated honestly where that is all we have, the label reference value. The reference value itself is kept for showing how a label's own dose compares against it, which is a separate question.

Where a product sits, and among what

Every product we can measure gets a position inside its comparison group. A group is one delivery format within one comparison scope: usually a single chemical form — zinc picolinate in tablets — but where the forms of an element differ only in how much of the element they carry, the scope is the element itself, as with creatine. Protein powders are compared within one protein source — whey concentrate with whey concentrate, pea with pea — because sources differ in amino-acid make-up, not only in how much protein they carry.

Inside the group we order products by unit price: euros per 1000 units of the element. Milligrams for most, micrograms where the doses in that catalogue are measured in micrograms; for protein powders, euros per 100 g of declared protein. Three equal parts: bottom, middle, top third.

Most products do not have a position, and that is deliberate. Today 1,405 of the 1,886 priced products we hold carry a unit price, and only part of those a position. A product stays visible in the catalogue without a position when it is a multi-element blend, when its label does not let us tell how much of the element is actually in it, when its delivery format is one we do not yet compare — a scoop and a 70 ml shot are not the same room as a tablet — or when the group is still too small to describe a distribution at all. Absence of a mark is not a verdict.

The comparison group is global. If you have set a shipping region, you are seeing part of it — the number on the mark says how many products the position was measured against, not how many are on your screen.

Why only the bottom third is marked

We mark the bottom third. We do not mark the middle or the top.

The reason is not technical. Praise and blame are not symmetric: “the cheaper third of this form” is a position statement that a wrong calculation makes embarrassing; “the more expensive third of this form”, attached to a named brand, is one that a wrong calculation makes harmful. The position is calculated either way. We simply do not turn the upper end into a mark.

Why units per day are on the card but not in the category

Products used to carry a second axis: how many units a day it takes to reach the target for that element. What that target is derived from, and why it is not the label reference value, is set out under Why the target dose is not the label reference value.

Measured on 25 August 2026: of the fourteen groups that carry a position, the axis was unusable in eight. In seven the median was below one unit a day — the middle of the group was splitting products between a fraction of a capsule and another fraction of a capsule, and nobody consumes a supplement that way. In the eighth no target figure exists at all. An axis that divides a quantity nobody uses, or that cannot be computed, is not a measurement.

The number itself stays on the card and it says something useful. If one serving covers the target several times over, that is a fact about the bottle: this product is dosed well above it. If it takes four capsules to get there, that is a fact too. We show the number and leave the judgement out — which is also why it carries no colour.

What the upper-limit note means

If, at the highest daily serving the label allows, the amount of the element exceeds the EFSA tolerable upper level, we say so. Which serving we use, and why the safety figure takes the top of the range while cost and duration take the midpoint, is set out under How we handle dosage ranges.

The warning has an edge, and it fails silent. Above three times the upper level we stop reading the serving instruction altogether — a label that extreme is more likely to be a label we have misread than a product anyone should take, and we will not build a safety claim on it.

No warning means we could not make the statement. It does not mean we checked and found it safe. Where an upper level applies to supplements only rather than to total intake, we compare against the supplemental figure alone; food is not counted.

4. Non-linear absorption

Absorption is not proportional to dose. For forms limited by active transport, the absorbed amount reaches a ceiling: beyond a certain single dose, taking more delivers no additional absorbed material, so the absorbed percentage falls.

We apply this ceiling only where a measured maximum absorbed amount exists in the literature for that specific chemical form, with a citation. Where no such measurement exists, we apply no penalty and no estimate: the coefficient is used as measured, and the product is neither rewarded nor punished for a saturation point we cannot source.

5. Data provenance — how our science is reviewed

Transparency is our only product, so we state plainly how each scientific value reaches our rankings — and how confident we are in it.

Every chemical form carries one of three evidence levels:

Unreviewed — the value is a working estimate and its science section is not published. Products may still appear in catalogue listings.

AI-reviewed — an AI agent has searched the published literature (primarily PubMed) and EFSA opinions, compiled the supporting studies with their PMIDs, and a second AI agent has cross-checked that compilation. Coefficients are set at the conservative end of the retrieved range. Reviews are refreshed quarterly, so new evidence reaches the rankings.

Human-audited — a named researcher has confirmed the value against the cited studies. This is the strongest level; we are working toward independent audit by a laboratory or academic partner.

What AI review means, honestly. AI-reviewed values are compiled from public research by an automated process, not by a qualified human researcher. AI can make mistakes — misread a study, conflate similar compounds, or miss relevant work. Every coefficient links to its source studies so you can check them yourself. If you find an error, tell us (see Limitations and corrections below).

Every claim is traceable. Each bioavailability coefficient and molecular fraction carries a PubMed (PMID) citation. Molecular fractions are calculated from molar mass and are verifiable arithmetic, not estimates.

For protein there is no molecular fraction. The protein figure is the manufacturer's declared value; in the EU it is calculated from nitrogen (N × 6.25). We do not measure it.

What we do not do. We do not recommend buying or avoiding any specific product, and we do not provide medical advice. Our health-related categories use only claims authorized under EU Regulation 1924/2006 at the element level. Form-specific research findings are shown as cited observations, never as health claims.

6. Limitations and corrections

Our model is deliberately strict, but nutrition science and human physiology are moving targets. We state the limits plainly:

Individual biology varies. Stomach pH, gut microbiome, and the presence of cofactors (such as vitamin D or vitamin C) all change real-world absorption. Our coefficients represent a conservative clinical average for healthy adults, not a personal prediction.
Baselines are conservative. Where the literature offers a range, we choose the cautious end. We would rather understate a benefit than overstate one.
We update on better evidence. When new peer-reviewed, ideally double-blind placebo-controlled pharmacokinetic data appears, we revise the baseline coefficient and the algorithm recalculates all affected rankings automatically.

Open to correction. If you believe a coefficient or rule is outdated or wrong, send us a peer-reviewed PubMed citation. Confirmed corrections are made publicly and logged in the changelog below. When the comparison layer itself is not defensible, we remove it instead of preserving a misleading badge.

What we cover, and what comes next

The method on this page is general; the catalogue is not yet. Today it covers the elements listed on the Elements page. An ingredient is added when three things exist for it: a comparison basis (elemental mass for minerals, declared protein for protein powders, a standardised marker compound for botanicals, colony-forming units for probiotics), its chemical forms, and a reviewed upper limit where one applies. Until then a product containing it stays visible in search with a sentence saying why it has no number, and carries none. The same page lists what is coming, without dates.

Protein powders

We price protein powders per 100 g of the protein the label declares, from the shelf price, the number of servings or the net weight (or volume), and the protein per serving or per 100 g (100 ml for liquids), as sold — not as prepared with milk. Where the label gives both, we check that they agree; if they differ by more than a tenth, we show no number. Where the label gives only the scoop weight, or mixes grams and millilitres, we show no number; we do not assume a density. Where a product mixes whey concentrate and isolate, we list it as concentrate; where it mixes protein sources, we show it without a rank.

Declared protein is calculated from nitrogen. Free amino acids such as glycine or taurine add nitrogen without being whole protein. Where the ingredient list shows them, we say so on the product. We do not adjust the number.

We compare within one protein source and publish no ranking across sources. Collagen is shown with its price but is not ranked: it contains no tryptophan, one of the nine indispensable amino acids (PMID 31096622).

There is no EU reference value or upper level for protein in supplements, so protein products carry no daily target, no units-a-day figure and no upper-limit note.

How we handle dosage ranges

Many labels give a range rather than a single serving: “take 1–2 capsules daily”. All of our per-day calculations use the midpoint of that range.

midpoint = (low + high) / 2

  "1–2 capsules"   →  1.5
  "2–4 capsules"   →  3
  "1–3 scoops"     →  2
  "2 capsules"     →  2      (no range — the number is used as given)

This is a declared convention, not an interpretation of the manufacturer's advice. We publish it before we use it, and it does not change based on the result. The range itself is always shown next to the number, so you can reproduce any figure from the label and our published constants.

A midpoint of 1.5 capsules is an arithmetic basis, not a dosing instruction. Where a label gives “1–2 daily”, we show both the range and the basis we used.

Three figures are derived from the midpoint:

element per day — elemental mg per serving × midpoint
how long the bottle lasts — units per container ÷ (units per serving × midpoint)
cost per 30 days — price ÷ days × 30

One figure is not. The upper tolerable intake (UL) flag uses the high end of the range, never the midpoint. If a label says “1–4 capsules” and four capsules exceed the EFSA upper limit, that is a fact you should see. Averaging it away would hide it. On safety, we always take the largest serving the label allows.

If the label gives no daily serving, we do not assume one. Cost per 30 days, bottle duration and element per day are left blank, and the product card says so. We do not default to one serving per day.

Changelog

2026-10-16 — Protein powders now carry a price per 100 g of declared protein, compared within one protein source. An earlier line on this page said protein would be compared by protein quality; we chose declared protein within one source instead, because quality scores depend on the reference pattern and are mostly measured in animals.
2026-08-25 — This page used to describe categories built on thresholds: a product was flagged when its cost per absorbed milligram reached three times the element's median, or when a form's absorption coefficient fell below a fixed cut-off. Those numbers were decisions, not measurements, and nobody could answer why three and not two. They have been removed. What replaces them is one measured position: where a product's unit price falls inside its own comparison group. A second axis — units per day — was tried and removed before it reached this page; the measurement that killed it is in the paragraph above.
2026-08-24 — For elements dosed in micrograms we show price per 1000 µg instead of per 1000 mg. The calculation is identical; only the scale differs. Which elements those are is our own editorial choice, recorded per element in the catalogue — not a threshold computed from the data.
2026-08-23 — Dosage ranges now use the midpoint of the label range in all per-day calculations (cost per 30 days, bottle duration, element per day). The range itself remains visible alongside every figure. The upper tolerable intake flag continues to use the high end of the range, not the midpoint. Sorting by labelled serving count has been removed: it ranked products by serving size rather than by convenience, which was a misleading reading. Two sorts replace it: element per day and bottle duration.
2026-08-18 — Within one chemical form, products are now ranked by euro per 1000 mg of the element, not by an absorption-adjusted score. Searching an element name groups results by form. A form page lists every product we carry in that form; products whose elemental amount is still uncertain stay on the list but are not named as cheapest and do not enter a percentile. Until the evidence review for a form is filled in, every form says we do not know — not that the form is fine, and not a blank. Labelled servings that would exceed three times the tolerable upper level are withheld rather than shown as a daily regimen.
2026-08-18 — The primary monthly cost on a product page is now the labelled serving, when the label states how many servings per day to take. That figure is days in the bottle and the 30-day cost at the manufacturer's recommended serving, including a range when the label gives one. The older scientific monthly cost (the cost of reaching the supplemental target) remains visible underneath and is not a ranking key. When one serving already covers the target, that scientific figure stays withheld; the labelled-serving cost takes its place. Coverage of that labelled field is 62% of products that have a scientific score; the rest keep a blank rather than a default of one serving per day. We also state, as two facts and without a verdict, the elemental amount at the labelled dose as a percentage of the EU reference, and that amount next to the tolerable upper level.
2026-08-18 — A monthly cost is no longer shown when one serving already covers the target. Roughly two in five products need less than one unit per day to reach the supplemental dose. The euro figure we used to show for those products priced a regimen that does not exist: a quarter of a capsule is not how anyone takes a supplement. We withhold that figure. What remains is a dose fact: one serving provides N× the target. Where the label states servings per day, the monthly cost of that labelled serving is now the primary figure on the product page.
2026-08-18 — Burden and monthly cost are now based on a supplemental target dose, not on the label reference value. For magnesium the two figures contradicted each other: the label reference value is 375 mg of total intake, while the tolerable upper level for supplemental magnesium is 250 mg, and the larger of the two was driving the calculation. Magnesium is now targeted at 250 mg, so the required daily units and the monthly cost of every magnesium product fall by roughly a third. Zinc, iron, vitamin D and omega-3 are unchanged in value, because for those elements the two figures never conflicted; what changed is that each target now records the basis it came from. One further consequence, stated plainly:
- Ten magnesium products that were withheld for requiring an implausible number of units per day now fall below that limit and are visible again. Their underlying uncertainty has not been resolved: for some of them we are not yet certain whether the label figure is the compound or the element. That question is being worked separately, and we chose not to hide ten products for a doubt that applies to several hundred.
2026-08-18 — Public category badges and value verdicts are being removed, not replaced. The following public surfaces no longer present whole-product verdict labels such as low-bioavailability, below-median-value, golden-standard, or similar market-comparison badges. We also removed public “save X% vs median” copy, homepage shame/champion blocks, and named hub leaderboards. This does not mean every product number changed. Some products simply lost a public interpretation layer while their underlying price or burden inputs stayed the same.
2026-08-11 — The saturation penalty now requires a measured absorption ceiling. It used to be triggered by the EFSA reference intake, which measures how small that reference is next to a normal serving — not whether the form actually saturates its transporter. Iron's reference is 14 mg and almost every iron supplement is 14–28 mg, so the penalty was firing on nearly the whole category for the wrong reason. Four consequences, stated plainly:
- 67 products' price-per-absorbed-amount improved, because the penalty no longer fires on the reference intake.
- 5 of those (ferrous sulfate heptahydrate, zinc oxide) improved by roughly 3×. They had been pinned to the model's absorption floor. At the time, their public comparison status did not improve just because the price-side number moved.
- The penalty returns to any form whose maximum absorbed amount is measured and cited. This ties the penalty to evidence; it does not abolish it.
- 4 iron products moved to a worse category without their own numbers changing at all. Our categories describe a product's position within its element, and the iron median fell 59% once 31 other iron products improved. Nothing about those 4 products got worse; the field around them got better. Market medians and category labels updated with the overnight refresh on 2026-08-12.
2026-08-04 — Element hub chemical-form citations now link each PMID token to its PubMed article. Citation wording (including provisional / animal-study caveats) is unchanged.
2026-07-23 — Introduced three-level evidence status (unreviewed / AI-reviewed / human-audited). Consolidated duplicate chemical form entries; magnesium coefficients revised toward the conservative end following literature review.
2026-07-17 — Initial public methodology published.

Public methodology note: the units-a-day figure remains public, while cross-form bio/value verdicts and median-based badges are being retired from user-facing surfaces. Last updated: October 2026.