Retail Reason-Why Standard
Last updated September 1, 2026
A reason-why sentence explains to a shopper why a product appeared in their results. It is published on a product page, alongside a commercial offer, in response to information the shopper gave us. That makes it a commercial communication.
Under the Unfair Commercial Practices Directive, the trader who publishes a commercial communication is answerable for it. A sentence that describes what a product does is a claim, and claims require evidence. A sentence that describes what is in a product is a fact, and facts require data.
This standard exists so that every sentence we generate is the second kind wherever it possibly can be, and so that where it cannot be, the sentence is placed and attributed correctly. It sets out nine commitments. Sections 5 and 6 are the ones that create obligations on both sides, and are the intended basis for the corresponding terms of service.
No sentence enters the string table unless it derives from a named source of truth.
There are five, and no others:
Ingredient classification — the INCI list, resolved against our Source, Group, Utility and Effect flags, plus the regulated-allergen lists.
Product category — what the product is: a powder, a lip stain, a scrub, an oil-based formula.
Label data — values printed on the pack and carried in the product feed, principally SPF.
Shopper input — the selections a person made in the diagnostic.
Our curation rules — published, versioned lists and bands we maintain and stand behind.
Each sentence in the library records which of these it draws on. A sentence whose origin cannot be named is not softened or hedged; it is removed. In the September 2026 review, four sentences failed this test and were withdrawn rather than reworded.
Composition, not effect
We describe what a product contains. We do not describe what it will do to a person. That is the difference between something evidenced from an ingredient list and something requiring a substantiation dossier held by the brand, not by the retailer.
An ingredient claim is not a product claim
Where a sentence names an ingredient function — soothing, hydrating, anti-dandruff — it says those ingredients are present. It does not say the finished product achieves that result. The distinction is explicit in EU guidance on cosmetic claims, and it is the reason these sentences are permissible at all.
The surrounding page must not close that gap. A composition sentence sitting under a heading that promises a result becomes a product claim by placement, and the heading is the retailer's.
No disease, no diagnosis
No sentence names a medical condition, and none describes a shopper's skin or hair in clinical terms. A product presented as treating or preventing a condition is a medicinal product by presentation and leaves the cosmetic framework entirely.
Acne, rosacea, dermatitis, alopecia, scarring and cellulite are all out of the library. Cosmetic vocabulary — blemish-prone, sensitive, dry, oily, fragile — stays, because it describes something the shopper told us rather than something we diagnosed.
No ingredient denigration
We do not present a legally permitted, safety-assessed ingredient as a defect. Free-from wording is never a badge and never sits next to any suggestion that an ingredient is unsafe. Where a shopper asked to avoid something, we report the match against their own filter — a fact about their request, not a judgement about the product.
Two rules that close the remaining gaps
No comparative without a stated basis. Better, better suited, not suitable, best for — all out. A comparative needs a defensible basis, and a retailer ranking one brand against another needs it in writing.
Never assert formulator intent. Chosen for, designed to, made for, formulated for — none of these are visible in an ingredient list. Only Contains, Listed as, We list this under, Matches your and We suggest.
How personalisation is expressed
We report what the shopper told us
A sentence never characterises a person's skin or hair. It repeats a value they entered. There is nothing to substantiate about a shopper's condition when the sentence does no more than read their own answer back to them.
A mismatch is a filter result, not a verdict
Negative sentences report that a product failed a filter the shopper set. They do not rank it, allege a defect, or compare it unfavourably with anything.
Advice is owned, and the rule behind it is published
Some sentences advise: lighter textures for blemish-prone skin, gentler exfoliants for sensitive skin, daily sun protection. Advice is lawful where it is visibly a recommendation rather than dermatological fact, where it is applied identically across every brand, and where the rule behind it is published.
The premise is never restated as science. We suggest oil-free textures is an editorial position. Oils cause breakouts would be a claim about human physiology, and we do not make it.
Data limits we work within
Three limits shape what a sentence can say. They are stated here because they explain wording that would otherwise look unnecessarily careful.
Presence, not concentration
An ingredient list says which ingredients are present, not how much of each. Ordering is descending only above one per cent; below that it carries no information. So a sentence never asserts a consequence that depends on level. Contains alcohol is true of a high-alcohol toner and of a trace preservative alike; may cause irritation would be reliably true of neither.
Absence statements fail false, so they are gated
Where classification data is incomplete, a positive statement under-reports; a negative statement becomes wrong. Contains parabens merely misses an entry. Contains no parabens is false — and false for exactly the shopper who set the filter.
Two things follow. Every block carrying an absence statement carries a scope line, which names the source rather than apologising for it:
And an absence statement does not render at all unless the data supports it: an ingredient list has been resolved, every ingredient on it matched our classification, and the record has been refreshed within our threshold. Composition statements are unaffected, so a gated card is shorter, never empty. The gate fires on not knowing, which is the only signal available — a product whose reformulated list contains one ingredient we cannot classify is precisely the case where an absence statement would have been wrong.
Ingredients are named twice
An ingredient has a name a shopper can read and a name they can check, and they are not interchangeable. Water and Aqua are one to one, but Vitamin C covers four different substances our classification treats separately, and Vitamin E covers two.
So a sentence carries the readable name in brackets, with a reference resolved at the foot of the block:
Ingredient names are never printed inline: several run long enough to take three or four lines on a phone and bury the sentence they were meant to support. They belong in the block line, where length costs nothing. Every ingredient a statement rests on appears there, including any left out of the bracket, so a shortened display never becomes a partial basis.
One rule governs the readable name. A derivative does not inherit its parent's name. Tocopherol is vitamin E, but Ascorbyl Glucoside is not vitamin C, Sodium Hyaluronate is not hyaluronic acid, and Retinyl Palmitate is not retinol. Calling them so is a statement about what the ingredient is, and would have to be substantiated like any other. Botanicals are the safe case, because there the common name is simply the plant.
Sentences doing compliance work — allergens, free-from wording, preference matches, a shopper's own block list — skip all of this and print the ingredient name directly, because there it already is the readable name. Someone avoiding Linalool is looking for the word Linalool.
Regulatory horizon
Two changes bear directly on retail reason-whys and are reflected in the current wording.
Directive (EU) 2024/825, applying from 27 September 2026, bans generic environmental claims without demonstrated excellent performance, uncertified sustainability labels, and presenting a legally required characteristic as a distinctive feature. Existing copy is not grandfathered, and the obligation falls on whoever publishes the page. Terms such as 100% natural and cruelty-free have been withdrawn or rebuilt as specific, sourced statements.
Regulation (EU) 2023/1545 extended the declarable fragrance allergen list to 82 substances, with labelling required from 31 July 2026. Our allergen wording refers to the regulated allergens we check for rather than naming a jurisdiction, so the sentences stay accurate as lists change and as further regulators — including the FDA under MoCRA — publish their own.
Both dates are live. The fragrance allergen deadline has passed; the green claims ban applies from 27 September 2026, and existing stock and copy are not exempt.