How Drop cites its sources

Not every citation is the same shape. Drop tags every entry in the database with one of five evidence-strength levels, and that tag is surfaced in-app on the flag itself — so you can tell at a glance whether a claim is anchored in a meta-analysis or in a single mechanistic study.

• Systematic review — the strongest tier. A published review that pools and weighs multiple primary studies, usually with a Cochrane-style methodology.
• RCT — randomized, controlled trial. The strongest single-study design for an effect claim.
• Observational — cohort or case-control studies. Useful and often the only thing available for skincare-adjacent topics, but a step below RCT for causal claims.
• Expert consensus— AAD practice guidance, Cosmetic Ingredient Review reports, dermatology textbook consensus. We surface these when primary studies do not exist or do not bear on the user’s question directly.
• Debated— we use this tag explicitly when the evidence is contested. If you see a “debated” tag in Drop, it means reasonable researchers disagree, and we tell you that rather than picking a side quietly.

Systematic review and RCT are the strongest tiers. The other three are still worth citing — but we surface the tier so you can weigh the claim accordingly.

We are deliberate about what we will and will not pull from. The bar is set so a citation in Drop means something specific.

• Peer-reviewed journals indexed by PubMed are the default source. Each entry records authors, year, journal, DOI, and PubMed ID where available.
• Preprints are cited only when no peer-reviewed source exists for the specific question, and we mark them as such.
• Manufacturer-funded studies are disclosed in the entry when we use one. We do use them sometimes — they are often the only studies on a specific formulation — but you will see the disclosure.
• Expert-consensus sources we draw on include AAD practice guidance, the Cosmetic Ingredient Review (CIR), Cochrane Reviews, and JAAD (Journal of the American Academy of Dermatology) clinical reviews.
• What we will not use:skincare blogs, brand marketing pages, “as seen on TV,” influencer claims, or affiliate advertorials. These never appear as citations in Drop.

The citation database is organized into nine categories (A through K, skipping I and J for clarity). Each category covers a specific bucket of questions we hear from users:

• A. Active ingredient conflicts — when stacking two actives raises irritation risk (retinol + an exfoliating acid such as glycolic acid (an AHA) the same evening, vitamin C + copper peptides, and so on).
• B. Active ingredient synergies — combinations that work better together than alone.
• C. Layering order rules — what goes on first, wait times, where the myths overstate the case.
• D. pH and concentration considerations — when an ingredient only works in a specific pH window, or where concentration matters more than presence.
• E. Pregnancy and breastfeeding — ingredients we flag for hard exclusion or caution during pregnancy.
• F. Age-specific guidance — under-18 caution flags, over-45 changes, age-band defaults.
• G. Skin condition cautions — when a known condition (eczema, rosacea, fungal acne, melasma, perioral dermatitis) changes the calculus and we route you toward a dermatologist.
• H. Effectiveness timelines — typical results windows so you can tell if a product is on track or quietly not working.
• K. K-beauty and Asian beauty — snail mucin, propolis, ginseng, mugwort, and the wider layering culture that the Western category system tends to miss.

Each entry records the same shape: an evidence-strength tier (above), the citations backing it (authors, year, journal, DOI, PubMed ID), what each citation specifically supports, a plain-English claim, an optional rationale, and a recommended action. The schema is published in our open-source repo.

Before any batch of new entries ships, we spot-check 10% of them by hand — pulling the cited paper, reading the abstract or the relevant section, and confirming the entry accurately reflects what the source says. The rubric is simple: does the citation actually support the claim, at the strength tier we labeled it?

Skin science is genuinely contested in places. We try to write like that is true, instead of like everything is settled.

• We do not use marketing-proof shorthand — phrases like “proven by science,” “proven in the clinic,” “recommended by doctors,” or “tested by dermatologists.” These constructs obscure how strong the evidence actually is, and they are usually doing marketing work, not evidence work.
• We do not say a cosmetic ingredient has been approved by the FDA. The FDA does not approve cosmetics — only over-the-counter drug ingredients and prescription products. Drop uses that phrasing only in the narrow OTC-drug context where it is accurate.
• We do not promise “guaranteed” results, a “perfect” routine, or “the” right answer. Skincare is highly individual; we surface the evidence and the typical case, and leave the choice to you.
• When the evidence is debated, we say so. The “debated” tag in the evidence hierarchy above exists for exactly this reason — there are real disagreements in the dermatology literature, and pretending otherwise would erode the thing that makes Drop useful.

We will be wrong sometimes. The database is curated by humans (with Claude assistance, disclosed in our build process), and curation at this scale produces mistakes — a citation that does not quite support the claim, an evidence tier that overstates the strength of a single study, a category gap we did not see.

If you spot one, please tell us. Pre-launch, corrections route to hello@drop-skincare.com; post-launch we open a dedicated corrections@drop-skincare.com inbox. We acknowledge corrections publicly when they change a published entry — transparency about what changed and why is part of how this stays honest.