Go back

Go back

Go back

Go back

La Mer Virtual Skin Analysis

La Mer Virtual Skin Analysis

Personalized, ML-assisted skincare consultation that standardizes recommendations and boosts attachment rate/AOV in prestige retail.

Personalized, ML-assisted skincare consultation that standardizes recommendations and boosts attachment rate/AOV in prestige retail.

+15%

Average order value

-30%

Time to recommendation

+31%

Attachment rate.

+15%

Average order value

-30%

Time to recommendation

+31%

Attachment rate

At a glance.

At a glance.

  • Role: Senior Product Designer (end-to-end)

  • Company: Estée Lauder Companies — La Mer

  • Timeline: 2023–2025

  • Team: Retail Ops, PM, iOS Eng, Data/ML, Brand, Research

  • Constraints: On-counter time box, privacy in-store, low-light conditions, multi-region locales

  • Platforms: iPadOS (in-store), handoff to mobile follow-ups

  • Tools: Figma, proto testing, analytics

  • Outcomes: Standardized consults; +15% AOV; higher associate confidence


Executive summary.

Executive summary.

  • Designed a guided, ML-assisted in-store consultation that standardizes La Mer recommendations at the counter.

  • Mapped the end-to-end retail journey (greeting → diagnosis → plan → checkout), reducing variance across consultants.

  • Introduced bundle logic and clearer rationale to increase attachment of complementary products.

  • Shipped a production iPad flow used by store associates; +15% AOV from bundled recs.

  • Created reusable patterns & governance later reused by other brands.

Problem & context

Problem & context

Luxury skincare consults were inconsistent and time-intensive. Associates relied on personal heuristics, leading to variability in recommendations, explanation quality, and basket size. We needed a guided, trustworthy experience that:

  • makes associates faster and more confident

  • grounds recommendations in visible signals and routine templates

  • gently increases attachment rate/AOV without feeling salesy.


Success criteria: increase bundle attachment, reduce idle/decision time, improve perceived professionalism and trust.

Research & key insights

Research & key insights

Methods: store shadowing, contextual inquiry, 1:1 interviews with associates and clients, desk research on consultation heuristics; later, moderated on-device prototype sessions.


What changed the design:


  • Explainability → trust. Customers commit more when they see why a product fits their concern (signal → rationale → product).

  • Time boxes are real. Typical consults run 8–12 minutes; any step >60s without progress feels long.

  • Attachment at “plan” time. Framing routines as AM/PM steps with clear roles (cleanse → treat → moisturize) improves bundle acceptance.

Validation & experiments

Validation & experiments

A/B concept tests (list vs. routine framing → routine wins on comprehension/attachment intent).


Copy tests for “Why this” (evidence-first vs. benefit-first). Week-long pilots in two stores tracking time-to-plan, Save/Share opt-in, and associate feedback.


Concept A/B: Routine framing outperformed a plain list on both comprehension (71% → 84%, +13pp) and attachment intent (44% → 58%, +14pp); n≈36, moderated.


Copy tests: “Why this” written evidence-first increased perceived trust (3.8 → 4.3, +0.5) and optional add-on selection (27% → 34%, +7pp).


Operational pilots (2 stores, 1 week): Time to plan dropped from 9m20s to 7m40s (−100s, −18%).


Attachment rate rose 44% → 56% (+12pp). Save/Share opt-in reached 41%. Associate confidence increased by +0.6 on a 5-point scale; plan completion improved 82% → 91% (+9pp).

Prototyping for A/B testing

Prototyping for A/B testing

A/B testing

A/B testing

A/B testing

A/B testing

Impact, risks & my role

Impact, risks & my role

Impact: +15% AOV from bundled recommendations (pilot basket mix vs. historical). Consistency up (standardized steps reduced variance). Associate confidence up (qualitative).
Risks & mitigations: over-prescription → optional clearly marked + diminishing-returns messaging; autonomy → substitutions and good/better/best presets; privacy → explicit consent, local processing where possible, clear deletion.


My role: led end-to-end IA, flows, interaction design, and pattern system; co-ran research synthesis & pilots with Research/Retail Ops; partnered with iOS & Data/ML on feasibility/explainability; authored governance for content updates.