E-commerce
June 28, 2026
"Is this serum suitable for my sensitive combination skin?" "Can I combine niacinamide and vitamin C?" "I am allergic to fragrances, what do you recommend?" On a DTC brand with 40 SKUs, these questions arrive before the purchase and after the first application.
Cosmetics decision support requires crossing skin type, concerns, INCI, and application order, without shifting into medical diagnosis. Cosmetics Business notes that a 6-question quiz coupled with personalized PDPs generated a +30% conversion rate on a catalog of 60 skincare products (Cosmetics Business, Skin AI 2026).
This guide #146 covers skin diagnosis, ingredients, allergies, and routine building for support and chatbots. Distinct from the general product quiz (quiz mechanics), complex products (#108) (transversal strategy), and regulated products (#119) (legal framework): here we focus on operational cosmetic advice.
Summary
Why is recommending cosmetics online more difficult than in-store?
In perfumery, the advisor sees the skin, asks three questions, and tests a texture. Online, the customer guesses based on a PDP and TikTok reviews.
DTC Cosmetics Specifics
Skin subjectivity: "sensitive" varies from person to person
Opaque INCI: Latin names, the customer does not read the list
Active interactions: retinol + AHA, mandatory SPF
Multi-product routine: 3 to 5 SKUs, not a single purchase
Return risk: skin reaction, wrong shade, product "does nothing"
The Cost of Poor Advice
RevenueHunt points out that a catalogue of 30 serums differentiated by skin type, active ingredient, and budget paralyzes the undecided: bounce or wrong purchase returned (RevenueHunt, skincare quiz 2026).
DTC Example
Clean skincare brand, 28% pre-purchase tickets asking "which product is right for me". 6-question diagnostic + INCI bot: skincare collection conversion +24%, "not suitable" returns -19%.
How does it differ from the general quiz and the regulatory framework?
Three complementary contents, three levels.
Product quiz (#guide quiz)
Product quiz: 5-7 questions mechanism, SKU mapping, zero-party data. This guide #146 details cosmetic content: which skin questions, which INCI filters, how to order a routine.
Regulated products (#119)
Regulated (#119): disclaimers, CAN/CANNOT, medical escalation. Here: authorized product advice within this framework, not a legal overhaul.
Supplements (#147)
See supplement selection help (#147): ingestion, EFSA claims, health advice limits.
What questions do customers ask about skin, ingredients, and routine?
Mapping out cosmetic questions feeds quizzes, bots, and support macros.
Skin Type and Concerns
« Oily vs. combination skin: which one to choose? »
« Dark spots, wrinkles, dehydration: where do I start? »
« Teenager vs. 50s: same range? »
« Reactive skin post-pregnancy »
Ingredients and Incompatibilities
« Does it contain fragrances / essential oils? »
« Compatible with the retinol I'm already using? »
« Non-comedogenic for acne-prone skin? »
« Vegan, cruelty-free, pregnancy-safe? »
Routine and Usage
Morning/evening order, frequency, SPF layering, bottle duration. See misunderstood products, pre-purchase objections.
How to structure a skin diagnosis in 5 to 7 questions?
An effective e-commerce skin diagnostic takes less than 60 seconds, with conditional logic.
Recommended sequence
Skin type: dry, oily, combination, normal, sensitive
Primary concern: hydration, radiance, anti-aging, blemishes, redness
Sensitivity: past reactions, self-reported rosacea (→ caution)
Allergies / exclusions: fragrance, essential oils, active ingredient X
Context: pregnancy/breastfeeding (→ disclaimer + handoff)
Current routine: beginner vs. retinol user
Budget / format: full routine vs. hero product
Conditional logic
Sensitive + fragrance excluded → filter tagged "fragrance-free" SKUs. Blemishes + beginner → avoid strong retinol, offer gentle salicylic acid. RevenueHunt: email capture at 50-70% quiz completion for nurturing without blocking results.
Zero-party data
Skin profile stored in Klaviyo: "dry skin + anti-aging" segments, moisturizing cross-sell follow-up. See zero-party data.
How do you explain the ingredients without overpromising?
The INCI explanation converts clean beauty customers if it remains factual and PDP-sourced.
Agent/Bot response rules
Cite official INCI from the product sheet, never from memory
Describe approved cosmetic role: "helps to hydrate", not "cures"
Refer back to the legally validated packaging claim
"Results may vary depending on skin type" systematically
Top active ingredients and safe formulation
Niacinamide: cite PDP concentration if displayed. Hyaluronic acid: cosmetic hydration. Retinol: progressive use according to instructions, daytime SPF, avoid during pregnancy. Vitamin C: form stability (derivatives) according to the product sheet. Inference Beauty decodes over 60,000 ingredients for customer transparency (Inference Beauty, beauty e-commerce).
Macro ingredient
"The [Product] contains [active ingredient] listed in the official INCI. It is presented for [approved cosmetic claim goal]. For medical questions or severe allergies, please consult a dermatologist."
How to manage allergies and ingredient exclusions?
The cosmetic allergy filter is a trust lever and return reducer if connected to structured data.
Common exclusions
Perfume / fragrance: `fragrance_free` tag
Essential oils: INCI pattern exclusion
Denatured alcohol: very dry skin
Lanolin, nut oils: declared allergy
Chemical vs mineral SPF: customer preference
Bot exclusion workflow
Customer declares "perfume allergy" → bot excludes any SKU with Perfume/Fragrance in INCI → recommends a short list of 2-3 products → 24h patch test reminder + dermatologist if severe history.
What you do not guarantee
Total absence of allergens without dedicated certification. Factory cross-contamination if undocumented. See regulated (#119), hallucinations (#123).
How do you build a routine based on slots rather than isolated products?
Recommending three incompatible serums is the classic mistake. The cosmetic routine slots structure the basket.
Routine slots day type
Cleanser: 1 SKU skin type
Treatment: target concern serum
Moisturizer: gel vs. cream texture
SPF: mandatory in the morning if photosensitive active ingredients are used
Evening routine slots
Double cleansing if wearing makeup, night active ingredient (retinol or AHA, not both for beginners), barrier cream. Bot suggests a complete routine of 4 slots max for beginners, not 8 intimidating products.
Incompatibilities to code
Retinol + AHA on the same evening = warning. Vitamin C + niacinamide: follow brand formulation (pH), refer to official guide. Haut.AI Skin.Chat adapts the routine in chat if the customer says "without retinol" (Haut.AI, Skin.Chat).
PDP Bundle
"Beginner mixed skin routine" kit pre-mapped quiz → 1-click basket. See product bundles.
How do I configure the cosmetics chatbot on Shopify?
The cosmetics chatbot combines PDP/INCI RAG, exclusion rules, and cautious handoff.
Minimum bot corpus
PDP: INCI, directions for use, authorized claims
Brand routine guide (order, frequency)
Active ingredients compatibility matrix validated by product manager
Shopify tags: skin_type, concern, fragrance_free
REG-COS macros disclaimers
Priority intents
`skin_type_match`, `ingredient_check`, `routine_builder`, `product_compare`, `usage_frequency`, `pregnancy_handoff`. Confidence gate 90 %+ on INCI; below threshold → "check product sheet" + agent.
Guardrails
Blocklist: cures, treats, prescribes, diagnoses. No recommendation without collected allergy exclusion. See automate (#120), bot limits (#124), bot product questions.
When to escalate to a human or a dermatologist?
Some questions go beyond commercial cosmetic advice.
Immediate Handoff
Ongoing skin reaction (redness, swelling, burning)
Pregnancy / breastfeeding / child < 3 years old question
Diagnosis request: "is this eczema?"
Medication + product interaction
Customer threatening DGCCRF report after agent claim
Escalation Script
"Your question goes beyond our role as product advisors. I recommend you consult a dermatologist. I can help you with the return according to our policy if you reacted to the product." REG-COS trained agent, no improvisation.
Human without medical emergency
Complex shade matching, existing 8-product routine to integrate, VIP customer, undecided cart > €150: product expert or brand-trained esthetician.
Which cosmetic consulting mistakes cost returns and negative reviews?
Five common beauty advice anti-patterns in support and bots.
1. Promising a time-framed result
"Spots cleared in 2 weeks" without an authorized claim. Disputes + reviews.
2. Invented or incomplete INCI
Bot hallucinates "fragrance-free" while Parfum is in the ingredients list. Allergy.
3. Overloaded routine
5 strong active ingredients recommended to a beginner. Irritation, return, brand blame.
4. Ignoring the existing routine
Customer already uses 1% retinol, bot suggests AHA + retinol on the same night.
5. Medical tone
"This product will treat your acne" vs "formulated for blemish-prone skin". See bot brand voice, quality of answers.
How does Qstomy help with choosing cosmetics?
Qstomy combines conversational diagnostics, INCI filtering, and PDP-anchored catalog recommendations.
Beauty Features
Skin diagnostic flow: 5-7 conditional questions
INCI exclusion: fragrance, flagged active ingredients
Routine slots: 3-4 compatible products
SKU comparison: texture, active ingredients, target skin type
REG Handoff: pregnancy, skin reaction
Add to cart: in-chat product cards
DTC Scenario in Figures
Skincare brand with 45 SKUs, 340 skin and ingredient questions/month. Qstomy deployment: diagnostics + INCI RAG + incompatibility matrix. After 5 months: pre-purchase advice tickets -46%, AOV routine bundle +€38, conversion of quiz/bot visitors +22%, mismatch returns -17%.
Explore AI support, Shopify, request a demo.
Which operational playbooks should be used to launch cosmetic product recommendation tools?
Playbook 1: Shopify skin tags (2 h)
Metafields: skin_type, concern, fragrance_free, pregnancy_caution. Apply top 20 SKUs. Sync bot corpus.
Playbook 2: 6-question diagnostic quiz (4 h)
Questions section 4, mapping slots, routine results page + add all. Links on homepage and PDP hero. Measure completion and conversion.
Playbook 3: incompatibility matrix (1 h)
Notion table: active ingredient A + active ingredient B → OK / avoid / alternate days. Validated by product manager. Integrated into bot.
Playbook 4: REG-COS macros (1 h)
5 macros: sensitive skin, pregnancy, reaction, INCI, comparison. Section 5 disclaimers. 30-min agent training.
Playbook 5: monthly audit of 20 conversations
INCI accuracy score, overpromising, consistent routine. 1 corpus fix/week.
Useful linking
Providing good cosmetic advice online does not mean playing the dermatologist: it means steering toward the right product, with the right words and verifiable sources.
Enzo
June 28, 2026
