On Lume’s Whole Body Deodorant Invisible Cream (scented), mandelic acid is the second ingredient — right after water and ahead of everything else. So the marketed hero is not buried; it is genuinely near the top. What the label will not tell you is the two numbers that decide whether that matters: the actual percentage of mandelic acid, and the finished pH. Both are absent from the box, and both are what published research says the outcome hinges on.
This is the site’s usual game — read the label, find where the star ingredient really sits — but Lume plays it back at us. Most “read the label” stories end with the hero ingredient hiding near the preservatives. Here it doesn’t. Mandelic acid is position 2. The interesting part is everything the ingredient order still can’t tell you.
Where does mandelic acid actually sit on the label?
Here is the full list, quoted verbatim from Lume’s own ingredients page for the scented Invisible Cream, read on 18 September 2026:
Water, Mandelic Acid, Maranta Arundinacea Root Powder, Tapioca Starch, Isoamyl Laurate, Hydroxypropyl Starch Phosphate, Cetearyl Alcohol, Caffeine, Caprylic/Capric Triglyceride, Behenyl Alcohol, Hydrogenated Castor Oil, Fragrance, Aloe Barbadensis Leaf Juice, Theobroma Cacao (Cocoa) Seed Butter, Floral Pyranol, Tocopherol, Allantoin, Panthenol, Sandalrome, Stearyl Alcohol, Dimethicone, Cetearyl Glucoside, Pentylene Glycol, Phenylpropanol, Potassium Hydroxide, Dipropylene Glycol, Sodium Benzoate, Tetrasodium Glutamate Diacetate
Source: lumedeodorant.com/ingredients (the unscented cream drops Fragrance, Floral Pyranol and Sandalrome; every other position is the same). Lume’s own page adds the caveat that the packaging is the authoritative list, so I’m analysing the scented tube exactly as published and naming it as such.
Ingredient lists in the US run in descending order of concentration down to 1%, after which order stops being required. Mandelic acid at position 2 sits well above that 1% floor, so it is present at a meaningful level — this is not the squalane-after-the-preservative trick. What follows are the cheap absorbent starches that give almost every “natural” deodorant its body — arrowroot powder and tapioca starch back to back, then a modified starch phosphate a few slots later. Under those, fatty alcohols and emollient esters. It’s a competent, ordinary natural-deodorant base built around one acid.
What can’t you read from the position alone?
Two things, and they’re the two that matter most.
The percentage. Position 2 tells you mandelic acid beats every ingredient except water. It doesn’t tell you whether that’s 3% or 12%. In the published work on mandelic acid’s antimicrobial activity, concentration is the whole story. A 2023 study of mandelic-acid hand gels (Łobacz et al., Int J Mol Sci) found the inhibition zones against odour-associated bacteria grew with dose — 10% did markedly more than 5%. A companion paper measuring the minimum inhibitory concentration (Szponar et al., Int J Mol Sci) put the MIC against Staphylococcus aureus near 0.175% in a test tube. So the bar to “does something” is low, but the effect keeps climbing with dose — and the label puts Lume’s number somewhere inside a very wide range you can’t see.
- What the label confirms
- Mandelic acid is the highest-concentration functional ingredient, present above 1%.
- What the label hides
- The exact percentage — the single figure the published dose-response work turns on.
- What the label hides
- The finished pH — which decides how much of that acid is in its active form.
Why does the potassium hydroxide near the end matter?
Second number: pH. An alpha hydroxy acid like mandelic acid only stays in its active, un-neutralised form in an acidic formula. Mandelic acid’s pKa is about 3.4, so at a pH around 3.5–4.0 most of it stays acidic; push the pH up and more of it converts to potassium mandelate, the salt. That conversion isn’t cosmetic: in the MIC paper above, the potassium salt was a weaker antimicrobial than the free acid.
Now look at position 25: Potassium Hydroxide. KOH is a base — it’s there to raise pH. That confirms the formula is being pH-adjusted, but not to what. Lume’s help pages describe a pH in the 3.5–4.0 range, which would keep most of the acid active if accurate; that figure isn’t printed on the box and a shopper can’t verify it. So the label confirms pH-tuning is happening and leaves you to trust the direction.
What about the 72-hour claim and the caffeine?
Lume’s headline has long been a clinically-tested 72-hour odour-control claim. The testing behind it was run by Princeton Consumer Research, a third-party consumer-testing centre, on a panel of women — a consumer study, not a peer-reviewed publication, and Lume’s own help page clarifies that 72-hour control doesn’t mean zero odour for 72 hours. That framing was challenged in Nelson v. Lume Deodorant, LLC (E.D.N.Y., 2023), which argued the study showed superiority over competitors rather than absolute all-day performance and that the samples weren’t large or diverse enough. That case was voluntarily dismissed with prejudice on 13 December 2023 — the plaintiff walked away, which is not a ruling either way. Separate 2025 suits naming Lume have since been reported; none has a certified class, and all allegations remain unproven. I’m pointing at the record, not the verdict.
There’s a regulatory reason a clinical-test claim reads differently here than on, say, a sunscreen. A mandelic-acid deodorant is a cosmetic; an aluminium-based antiperspirant is an OTC drug under the FDA’s antiperspirant monograph. Cosmetic claims don’t carry the drug-level substantiation bar, so a clinical-test claim on a deodorant means whatever the brand’s own testing says it means. The caffeine at position 8 is its own small puzzle: it has vasoconstrictive properties but no published mechanism for controlling underarm odour, and Lume’s materials don’t explain what it’s doing there. Read it as a label ingredient until someone shows otherwise.
Is the fragrance worth a second look?
The scented cream lists “Fragrance” at position 12, plus two named synthetic fragrance molecules further down (Floral Pyranol and Sandalrome). The underarm is one of the more reactive spots for fragrance contact allergy: a classic provocation study using deodorants (Johansen et al., J Am Acad Dermatol, 2003) showed that fragrance allergens at everyday concentrations could provoke axillary dermatitis in already-sensitised people within weeks, and the EU’s scientific committee has documented deodorants as a common trigger in fragrance-sensitive patients (SCCP opinion on fragrance allergy). That’s a risk for the fragrance-sensitised minority, not everyone — and “Fragrance” as an umbrella term means the specific allergens in this cream aren’t disclosed, so I can’t tell you which are present. The unscented cream sidesteps the whole question by leaving fragrance out.
So what would a simpler label look like?
If the appeal of Lume is “fewer moving parts than an antiperspirant,” it’s worth seeing what a genuinely short list reads like. Our own tallow deodorant — Ingredient Desk is published by the team behind Evil Goods, and I’ll always say so — is a tallow-based aluminium-free deodorant whose named ingredients you can count on one hand. I won’t quote a fixed number, because the site’s own copy isn’t fully consistent yet and I’d rather name ingredients than count them; the point isn’t that it’s better at odour control — that would need its own testing — but that a shorter, named list gives a reader less to wonder about. Lume’s list isn’t sinister; it’s just long, and two of the numbers that would let you judge it aren’t on the box.
The honest summary
Mandelic acid at position 2 is a real, near-the-top inclusion — credit where it’s due, no burial here. The formula’s bulk is water, starch-family absorbers and fatty alcohols: a standard natural-deodorant base, not a house of horrors. The fair critique isn’t “this is bad,” it’s “the label can’t answer the two questions that decide whether the hero ingredient is working” — its percentage and the finished pH — and the flagship clinically-tested 72-hour claim rests on a consumer panel that was pointed enough to draw a lawsuit before the plaintiff dropped it. Read the label, note what’s missing, and keep your expectations sized to what’s actually disclosed.
