ScentMatch
AI-powered fragrance and beauty product similarity search that helps consumers discover relevant alternatives and personalized matches across large e-commerce catalogs when keyword search falls short.
The Problem
“ScentMatch: AI-powered fragrance and beauty similarity search and personalized discovery”
Organizations face these key challenges:
Customers cannot smell products online and often struggle to articulate preferences
Keyword search fails for descriptive intent, mood-based queries, and inconsistent product naming
Visual browsing online does not replicate in-store discovery behavior
Catalogs contain many similar products with subtle differences in notes, concentration, and price
Out-of-stock items create dead ends without strong substitute recommendations
Customer journeys are fragmented across digital, print, in-person, and advisor-led channels
Recommendation quality suffers when relying only on rules or sparse behavioral data
Impact When Solved
The Shift
Human Does
- •Review catalog content and identify products that need manual substitute or similar-item mappings
- •Create and update keyword tags, category filters, and brand-based recommendation rules
- •Curate similar-product lists for key items, promotions, and out-of-stock substitutions
- •Check search and product page performance and adjust merchandising rules when results look weak
Automation
- •Return matches based on exact keywords, filters, and basic category logic
- •Surface recommendations from simple brand rules or collaborative filtering patterns
- •Flag zero-result searches and low-engagement queries in standard reporting
- •Apply existing substitute mappings and fallback rules when products are unavailable
Human Does
- •Set similarity goals, merchandising priorities, and guardrails for acceptable matches
- •Review and approve high-impact substitute recommendations, clusters, and merchandising changes
- •Handle exceptions involving brand conflicts, sensitive ingredient concerns, or poor-fit results
AI Handles
- •Analyze product descriptions, reviews, notes, ingredients, and metadata to identify similar items across the catalog
- •Generate and rank relevant alternatives, dupes, and cross-sell recommendations for search and product pages
- •Adapt recommendations using shopper context such as preferences, budget, and prior browsing behavior
- •Monitor catalog changes, detect weak similarity coverage, and propose updated substitute mappings and discovery opportunities
Operating Intelligence
How ScentMatch runs once it is live
AI runs the first three steps autonomously.
Humans own every decision.
The system gets smarter each cycle.
Who is in control at each step
Each column marks the operating owner for that step. AI-led actions sit above the divider, human decisions and feedback loops sit below it.
Step 1
Assemble Context
Step 2
Analyze
Step 3
Recommend
Step 4
Human Decision
Step 5
Execute
Step 6
Feedback
AI lead
Autonomous execution
Human lead
Approval, override, feedback
AI handles assembly, analysis, and execution. The human gate sits at the decision point. Every cycle refines future recommendations.
The Loop
6 steps
Assemble Context
Combine the relevant records, signals, and constraints.
Analyze
Evaluate options, risk, and likely outcomes.
Recommend
Present a ranked recommendation with supporting rationale.
Human Decision
A human accepts, edits, or rejects the recommendation.
Authority gates · 1
The system must not publish high-impact substitute mappings, merchandising changes, or expert-assisted scent recommendations without review by a merchandiser, category manager, or beauty advisor. [S2][S5]
Why this step is human
The decision carries real-world consequences that require professional judgment and accountability.
Execute
Carry out the approved action in the operating workflow.
Feedback
Outcome data improves future recommendations.
1 operating angles mapped
Operational Depth
Technologies
Technologies commonly used in ScentMatch implementations:
Key Players
Companies actively working on ScentMatch solutions:
Real-World Use Cases
Multichannel product recommendations across Avon sales channels
Avon used AI to suggest products a shopper is likely to want next, whether they interact through representatives, online stores, or digital brochures.
Looking Similar image-based product recommendations
The system looks at a product photo and finds other catalog items that visually resemble it, like helping a shopper say 'show me more things that look like this.'
Expert-assisted omnichannel fragrance matching using quiz, guides, and in-store sampling
Ulta helps you narrow options online, then encourages you to test scents in store and ask beauty advisors before buying.
Natural-language fragrance recommendation via semantic search
A shopper can describe the kind of scent they want in plain English, and the system finds perfumes/colognes with a similar vibe even if the exact words do not appear in product descriptions.
Fragrance Finder interactive scent quiz
A quiz asks about what scents you like and then suggests perfumes that fit your taste.