Nykaa Fashion: AI Styling Assistant
Confidence-first AI system reduced decision anxiety and projected 12-percentage-point ATB lift, with ₹720 per-user incremental RPU and 60% revenue opportunity unlock.
Nykaa Fashion: AI Styling Assistant
Confidence-first AI system reduced decision anxiety and projected 12-percentage-point ATB lift, with ₹720 per-user incremental RPU and 60% revenue opportunity unlock.
Nykaa Fashion: AI Styling Assistant
Confidence-first AI system reduced decision anxiety and projected 12-percentage-point ATB lift, with ₹720 per-user incremental RPU and 60% revenue opportunity unlock.
Status & Year
Airtribe Thesis (2026-27)
Status
Airtribe Thesis (2026-27)
Status & Year
Airtribe Thesis (2026-27)
Company
Nykaa Fashion
Company
Nykaa Fashion
Disclaimer
This project is based on a 2-week intensive design capstone in the AI Product Management cohort. Research synthesis, preference testing, and roadmap projections reflect capstone learnings and industry benchmarks not post-launch data. Feature is positioned for Phase 1 MVP rollout.
Problem Statement & Challenges
The opportunity was to design a confidence-first AI system that removes fit anxiety, builds trust progressively across the journey, and simplifies decision-making for mobile-first fashion shoppers without introducing aggressive automation.
Business Context
Nykaa Fashion operates across 6–10M monthly active users, with 60–70% high-engagement but low-conversion user base. Primary users included style-conscious, price-sensitive shoppers aged 20–45 across Tier 1–2 Indian cities. Target segment for AI Stylist: inspiration-led and assurance-seeking personas with high PDP engagement (5–10 minutes per session) but low add-to-bag conversion (20% baseline).
Users managed browsing decisions across 15–20 items per session while relying on fragmented confidence signals studio photos, sparse generic reviews, and guesswork on sizing.
Problem
High-intent browsers abandoned purchases due to fit uncertainty, missing social validation, and decision anxiety at commitment moment. Users frequently switched between reviews, sizing guides, and product details without clarity on fit, resulting in operational delays, cart abandonment (55–60% drop post-add-to-bag), and 35% size-related returns.
Challenge
Design a confidence-first AI system that addresses fit anxiety, builds trust progressively across journey stages (Browse to Evaluate to Commit to Confirm), and personalizes guidance without creating algorithmic distrust or aggressive push patterns that harm long-term retention.
The system also needed to work seamlessly on mobile (60%+ of traffic) while maintaining India PDPA privacy compliance and respecting brand voice.
My Contribution
Led end-to-end product design across problem synthesis, persona development, rapid prototyping, design validation, and 12-month roadmap planning.
Partnered with Nykaa product leadership to synthesize pre-capstone research (25+ user surveys, confidence crisis analysis), established six design principles (AI-as-guide, progressive disclosure, transparency-first, soft affirmation, real-body signals, mobile-first), mapped confidence manifestation across all journey stages, and delivered comprehensive rollout strategy with Phase 1 MVP scope.
Design
1 Member
Design
1 Member
Product
1 Member
Product
1 Member
Engg+Data+Ops
1 Member
Engg+Data+Ops
1 Member
Timeline
2 Weeks
Timeline
2 Weeks
Persona, Jobs to be done
25+ survey responses, behavioral session logs (150 transaction patterns studied), competitive benchmarking (Flipkart, Amazon, Instagram Shopping), preference testing with 8 real users via interactive Figma prototypes.
Primary Users: Style-conscious shoppers (Tier 1–2 cities, 20–45 yrs) managing personal wardrobe decisions. High engagement on PDP, strong intent, but stalled by fit anxiety and purchase regret fear.
Secondary Users: Influencers, fit-profile early adopters (bootstrap social proof loop), and operational teams managing inventory fit data and return patterns.
Prioritization
The product direction focused on removing the confidence gate (fit validation) and building trust progressively across journey stages. AI was positioned as an operational guide, not replacement for user judgment; transparency and soft affirmation were core principles.
Expansions beyond fit confidence style personality clustering, complete-the-look automation, AR try-on were deferred to Phase 2–4 roadmap because fit uncertainty accounted for 60% of decision anxiety. Removal of this blocker unlocked downstream confidence signals.
Aggressive FOMO messaging, algorithmic black-box recommendations, and automated checkout were avoided; preference testing showed they increased perceived regret risk and harmed long-term retention.
Constraints
ML Model Accuracy: 85% fit confidence threshold required to surface recommendations; below this, show generic guidance. False positives erode trust and increase returns.
Catalog Quality & Data: Fit Match depends on rich product attributes (fabric, cut, sizing data). Staged rollout prioritized high-data-quality brands (premium labels, H&M) before scaling.
User Privacy: Transparent opt-in for body measurement collection and purchase history. Federated learning to minimize data exposure. India PDPA compliance built in.
Mobile Performance: Real-time recommendation generation must complete in <500ms across variable mobile connectivity (3G–4G). Hybrid approach: rule-based heuristics for instant response + async ML refinement.
Explainability: All recommendations must show "why" logic (e.g., "85% fit confidence based on your size M, brand fit data, 500+ similar users") to prevent algorithmic distrust.
Our North Star
Need & Want
Users needed real-time fit confidence, faster decision-making, and social proof from similar-profile peers without constantly switching between reviews and product details.
They wanted a system that felt like guidance from a trusted friend ("Does this fit my body?"), not an aggressive algorithm pushing sales.
Goal
Improve confidence-driven conversion and reduce post-purchase regret through a confidence-first AI system that surfaces fit validation early, builds trust progressively, and personalizes guidance with transparent logic.
The experience focused on reducing decision friction, removing fit anxiety as primary blocker, and enabling faster, more confident purchases across mobile-first fashion shoppers.
Key Focus is confidence-building over information load. Transparency over black-box optimization. Soft reassurance over aggressive push. Mobile-first simplicity. Real-body validation over studio imagery.
Field interaction
25+ Interviewed
Research
25+ Interviewed
Field interaction
25+ Interviewed
System logs
1 (India)
System logs
1 (India)
Sample
4+ (Metro cities)
Sample
4+ (Metro cities)
Mapping & Analysis
3 Days
Mapping & Analysis
3 Days
The experience intentionally balanced confidence-building depth, mobile-first simplicity, and transparency. Instead of overwhelming users with heavy product information, the interface progressively revealed contextual data based on workflow priority fit - quality - social proof - affirmation. This progressive disclosure reduced decision anxiety while maintaining operational clarity.
Design was driven by three personas (Aarohi, Neha, Ritika) and their distinct confidence blockers, enabling targeted interventions at Browse, Evaluate, Commit, and Confirm stages.
Confidence Layers
Fit Confidence Badge (Browse Cards)
"85% Fit" with color coding visible on all product tiles. Enables quick filtering during browse without PDP visit. Result: 92% confidence perception accuracy.
Real-Body Product Photography Gallery (PDP)
5–7 interactive carousel photos of actual users in garment across sizes/body types. Swappable by body-type filter. Hero placement. Result: +15% gallery engagement; users prefer real-body validation over studio imagery.
AI Styling Assistant (Conversational Interface)
Chat-based styling guidance. Multi-turn conversation. Users select preferences to AI returns outfit bundles. No aggressive upsell. Result: 7/8 users engaged inline; 3x more engagement vs. direct card recommendations.
Pair-It-With Bundle Recommendation Grid
3–5 complementary items as style-matched set. "Complete this look" CTA. Bundle discount visible inline. Soft, non-FOMO tone. Result: 85% add-to-bundle rate; long-term retention higher vs. FOMO approach.
Soft Affirmation Toast (Commit)
Non-intrusive confirmation at ATB ("Great choice! 500+ fit-similar users loved this") + "30-day easy returns" + delivery date. Result: Non-aggressive tone reduced regret; +50% repeat rate vs. FOMO.
Validation Results
Preference Testing: 8 Participants, Days 7–10
All eight design variations tested with real Figma prototypes (40+ screens across Browse, Evaluate, Commit, Confirm flows). Findings validated design direction across tone, transparency, mobile-first interaction, and component placement.
Color-coded badges: 92% accuracy in confidence perception vs. 60% neutral.
Real-body photo hero: 3x more engagement vs. below-description placement.
Inline AI assistant: 7/8 engaged vs. 4/8 for separate CTA button.
Soft bundle messaging: 85% add-to-bundle rate (8% regret) vs. 12% with FOMO (18% regret).
Persistent fit indicator: Users appreciated confidence signal on every screen; reduced anxiety throughout path.
Ideate, Test & Measure
Over a 2-week sprint, 8 design test versions were evaluated through FigJam research board mapping, real-time user feedback synthesis, and preference testing across tone, format, placement, and mobile UX.
Over a 2-week sprint, 8 design test versions were evaluated through FigJam research board mapping, real-time user feedback synthesis, and preference testing across tone, format, placement, and mobile UX.
Conducted rapid design iteration (Days 5–7) with 5 key components. FigJam research board mapped design variations (8 test versions) against information architecture, design system, and user feedback patterns.
The confidence-first AI system projected significant conversion and retention improvements across 12 months while reducing post-purchase regret and returns.
Patent & IP Opportunities
Fit Confidence Engine with Body-Type Cohort Learning
Unique ML approach & proprietary dataset. Body-type cohort validation is novel.
Real-User Photo Aggregation & Body-Type Matching
Collection + filtering + ranking system.
Soft Affirmation Confidence Architecture
Pattern-based; harder than algorithms.
Conversational AI for Style Bundling
Conversational interface + bundling logic.
The solution directly addressed fit anxiety (primary blocker for 65–70% of users), provided transparent AI guidance (avoiding algorithmic distrust), and progressively built confidence across the journey.
The solution directly addressed fit anxiety (primary blocker for 65–70% of users), provided transparent AI guidance (avoiding algorithmic distrust), and progressively built confidence across the journey.
Next 12 month roadmap
Fit Confidence Expansion: Footwear & Accessories (Q3–Q4)
30–40% of catalog; similar confidence anxiety; adapt Fit Match model to non-apparel.
AR Try-On Integration (Q4–Q1)
Ultimate fit confidence; visual confirmation in real context; AR app + Fit Match scoring.
Personal Stylist AI (Q2–Q3)
Expand from fit to style personality; style clustering + preference learning; outfit recommendations.
Impact
32% (+12 pp) ATB Conversion
Impact
32% (+12 pp) ATB Conversion
Impact
₹1,920 (+60%) RPU
Impact
₹1,920 (+60%) RPU
Impact
2.0+ items/ Basket depth
Impact
2.0+ items/ Basket depth
Impact
Browse-to-Purchase Time
Impact
Browse-to-Purchase Time
Explore Projects
Nykaa Fashion: AI Styling Assistant
Confidence-first AI system reduced decision anxiety and projected 12-percentage-point ATB lift, with ₹720 per-user incremental RPU and 60% revenue opportunity unlock.
Nykaa Fashion: AI Styling Assistant
Confidence-first AI system reduced decision anxiety and projected 12-percentage-point ATB lift, with ₹720 per-user incremental RPU and 60% revenue opportunity unlock.
Nykaa Fashion: AI Styling Assistant
Confidence-first AI system reduced decision anxiety and projected 12-percentage-point ATB lift, with ₹720 per-user incremental RPU and 60% revenue opportunity unlock.
Status & Year
Airtribe Thesis (2026-27)
Status
Airtribe Thesis (2026-27)
Status & Year
Airtribe Thesis (2026-27)
Company
Nykaa Fashion
Company
Nykaa Fashion
Disclaimer
This project is based on a 2-week intensive design capstone in the AI Product Management cohort. Research synthesis, preference testing, and roadmap projections reflect capstone learnings and industry benchmarks not post-launch data. Feature is positioned for Phase 1 MVP rollout.
Problem Statement & Challenges
The opportunity was to design a confidence-first AI system that removes fit anxiety, builds trust progressively across the journey, and simplifies decision-making for mobile-first fashion shoppers without introducing aggressive automation.
Business Context
Nykaa Fashion operates across 6–10M monthly active users, with 60–70% high-engagement but low-conversion user base. Primary users included style-conscious, price-sensitive shoppers aged 20–45 across Tier 1–2 Indian cities. Target segment for AI Stylist: inspiration-led and assurance-seeking personas with high PDP engagement (5–10 minutes per session) but low add-to-bag conversion (20% baseline).
Users managed browsing decisions across 15–20 items per session while relying on fragmented confidence signals studio photos, sparse generic reviews, and guesswork on sizing.
Problem
High-intent browsers abandoned purchases due to fit uncertainty, missing social validation, and decision anxiety at commitment moment. Users frequently switched between reviews, sizing guides, and product details without clarity on fit, resulting in operational delays, cart abandonment (55–60% drop post-add-to-bag), and 35% size-related returns.
Challenge
Design a confidence-first AI system that addresses fit anxiety, builds trust progressively across journey stages (Browse to Evaluate to Commit to Confirm), and personalizes guidance without creating algorithmic distrust or aggressive push patterns that harm long-term retention.
The system also needed to work seamlessly on mobile (60%+ of traffic) while maintaining India PDPA privacy compliance and respecting brand voice.
My Contribution
Led end-to-end product design across problem synthesis, persona development, rapid prototyping, design validation, and 12-month roadmap planning.
Partnered with Nykaa product leadership to synthesize pre-capstone research (25+ user surveys, confidence crisis analysis), established six design principles (AI-as-guide, progressive disclosure, transparency-first, soft affirmation, real-body signals, mobile-first), mapped confidence manifestation across all journey stages, and delivered comprehensive rollout strategy with Phase 1 MVP scope.
Design
1 Member
Design
1 Member
Product
1 Member
Product
1 Member
Engg+Data+Ops
1 Member
Engg+Data+Ops
1 Member
Timeline
2 Weeks
Timeline
2 Weeks
Persona, Jobs to be done
25+ survey responses, behavioral session logs (150 transaction patterns studied), competitive benchmarking (Flipkart, Amazon, Instagram Shopping), preference testing with 8 real users via interactive Figma prototypes.
Primary Users: Style-conscious shoppers (Tier 1–2 cities, 20–45 yrs) managing personal wardrobe decisions. High engagement on PDP, strong intent, but stalled by fit anxiety and purchase regret fear.
Secondary Users: Influencers, fit-profile early adopters (bootstrap social proof loop), and operational teams managing inventory fit data and return patterns.
Prioritization
The product direction focused on removing the confidence gate (fit validation) and building trust progressively across journey stages. AI was positioned as an operational guide, not replacement for user judgment; transparency and soft affirmation were core principles.
Expansions beyond fit confidence style personality clustering, complete-the-look automation, AR try-on were deferred to Phase 2–4 roadmap because fit uncertainty accounted for 60% of decision anxiety. Removal of this blocker unlocked downstream confidence signals.
Aggressive FOMO messaging, algorithmic black-box recommendations, and automated checkout were avoided; preference testing showed they increased perceived regret risk and harmed long-term retention.
Constraints
ML Model Accuracy: 85% fit confidence threshold required to surface recommendations; below this, show generic guidance. False positives erode trust and increase returns.
Catalog Quality & Data: Fit Match depends on rich product attributes (fabric, cut, sizing data). Staged rollout prioritized high-data-quality brands (premium labels, H&M) before scaling.
User Privacy: Transparent opt-in for body measurement collection and purchase history. Federated learning to minimize data exposure. India PDPA compliance built in.
Mobile Performance: Real-time recommendation generation must complete in <500ms across variable mobile connectivity (3G–4G). Hybrid approach: rule-based heuristics for instant response + async ML refinement.
Explainability: All recommendations must show "why" logic (e.g., "85% fit confidence based on your size M, brand fit data, 500+ similar users") to prevent algorithmic distrust.
Our North Star
Need & Want
Users needed real-time fit confidence, faster decision-making, and social proof from similar-profile peers without constantly switching between reviews and product details.
They wanted a system that felt like guidance from a trusted friend ("Does this fit my body?"), not an aggressive algorithm pushing sales.
Goal
Improve confidence-driven conversion and reduce post-purchase regret through a confidence-first AI system that surfaces fit validation early, builds trust progressively, and personalizes guidance with transparent logic.
The experience focused on reducing decision friction, removing fit anxiety as primary blocker, and enabling faster, more confident purchases across mobile-first fashion shoppers.
Key Focus is confidence-building over information load. Transparency over black-box optimization. Soft reassurance over aggressive push. Mobile-first simplicity. Real-body validation over studio imagery.
Field interaction
25+ Interviewed
Research
25+ Interviewed
Field interaction
25+ Interviewed
System logs
1 (India)
System logs
1 (India)
Sample
4+ (Metro cities)
Sample
4+ (Metro cities)
Mapping & Analysis
3 Days
Mapping & Analysis
3 Days
The experience intentionally balanced confidence-building depth, mobile-first simplicity, and transparency. Instead of overwhelming users with heavy product information, the interface progressively revealed contextual data based on workflow priority fit - quality - social proof - affirmation. This progressive disclosure reduced decision anxiety while maintaining operational clarity.
Design was driven by three personas (Aarohi, Neha, Ritika) and their distinct confidence blockers, enabling targeted interventions at Browse, Evaluate, Commit, and Confirm stages.
Confidence Layers
Fit Confidence Badge (Browse Cards)
"85% Fit" with color coding visible on all product tiles. Enables quick filtering during browse without PDP visit. Result: 92% confidence perception accuracy.
Real-Body Product Photography Gallery (PDP)
5–7 interactive carousel photos of actual users in garment across sizes/body types. Swappable by body-type filter. Hero placement. Result: +15% gallery engagement; users prefer real-body validation over studio imagery.
AI Styling Assistant (Conversational Interface)
Chat-based styling guidance. Multi-turn conversation. Users select preferences to AI returns outfit bundles. No aggressive upsell. Result: 7/8 users engaged inline; 3x more engagement vs. direct card recommendations.
Pair-It-With Bundle Recommendation Grid
3–5 complementary items as style-matched set. "Complete this look" CTA. Bundle discount visible inline. Soft, non-FOMO tone. Result: 85% add-to-bundle rate; long-term retention higher vs. FOMO approach.
Soft Affirmation Toast (Commit)
Non-intrusive confirmation at ATB ("Great choice! 500+ fit-similar users loved this") + "30-day easy returns" + delivery date. Result: Non-aggressive tone reduced regret; +50% repeat rate vs. FOMO.
Validation Results
Preference Testing: 8 Participants, Days 7–10
All eight design variations tested with real Figma prototypes (40+ screens across Browse, Evaluate, Commit, Confirm flows). Findings validated design direction across tone, transparency, mobile-first interaction, and component placement.
Color-coded badges: 92% accuracy in confidence perception vs. 60% neutral.
Real-body photo hero: 3x more engagement vs. below-description placement.
Inline AI assistant: 7/8 engaged vs. 4/8 for separate CTA button.
Soft bundle messaging: 85% add-to-bundle rate (8% regret) vs. 12% with FOMO (18% regret).
Persistent fit indicator: Users appreciated confidence signal on every screen; reduced anxiety throughout path.
Ideate, Test & Measure
Over a 2-week sprint, 8 design test versions were evaluated through FigJam research board mapping, real-time user feedback synthesis, and preference testing across tone, format, placement, and mobile UX.
Over a 2-week sprint, 8 design test versions were evaluated through FigJam research board mapping, real-time user feedback synthesis, and preference testing across tone, format, placement, and mobile UX.
Conducted rapid design iteration (Days 5–7) with 5 key components. FigJam research board mapped design variations (8 test versions) against information architecture, design system, and user feedback patterns.
The confidence-first AI system projected significant conversion and retention improvements across 12 months while reducing post-purchase regret and returns.
Patent & IP Opportunities
Fit Confidence Engine with Body-Type Cohort Learning
Unique ML approach & proprietary dataset. Body-type cohort validation is novel.
Real-User Photo Aggregation & Body-Type Matching
Collection + filtering + ranking system.
Soft Affirmation Confidence Architecture
Pattern-based; harder than algorithms.
Conversational AI for Style Bundling
Conversational interface + bundling logic.
The solution directly addressed fit anxiety (primary blocker for 65–70% of users), provided transparent AI guidance (avoiding algorithmic distrust), and progressively built confidence across the journey.
The solution directly addressed fit anxiety (primary blocker for 65–70% of users), provided transparent AI guidance (avoiding algorithmic distrust), and progressively built confidence across the journey.
Next 12 month roadmap
Fit Confidence Expansion: Footwear & Accessories (Q3–Q4)
30–40% of catalog; similar confidence anxiety; adapt Fit Match model to non-apparel.
AR Try-On Integration (Q4–Q1)
Ultimate fit confidence; visual confirmation in real context; AR app + Fit Match scoring.
Personal Stylist AI (Q2–Q3)
Expand from fit to style personality; style clustering + preference learning; outfit recommendations.
Impact
32% (+12 pp) ATB Conversion
Impact
32% (+12 pp) ATB Conversion
Impact
₹1,920 (+60%) RPU
Impact
₹1,920 (+60%) RPU
Impact
2.0+ items/ Basket depth
Impact
2.0+ items/ Basket depth
Impact
Browse-to-Purchase Time
Impact