The problem involves the lack of personalization on Nykaa Fashion's discovery surfaces for logged-in users. A static editorial experience leads to decision fatigue when presented with thousands of undifferentiated products.
- User pain level: High. Returning users already have established preferences but must repeatedly apply filters on every session, wasting time.
- Frequency: Every session for a returning user.
- Existing workarounds: Manual filtering, searching for specific brands, or bouncing to competitors with better discovery feeds.
- Existing products: Myntra (adaptive category ordering), Sephora (re-ranking by profile), Purplle (profile onboarding + "For You" shelves). Big players like Amazon use heavy collaborative filtering.
- Strengths of competitors: Profile-driven curation reduces time-to-cart.
- Weaknesses of competitors: Often prioritize past purchases over current in-session intent (e.g., browsing for a wedding guest dress vs. usual casual wear).
- Unserved gaps: Real-time intent weighted against historical affinity. Fashion requires capturing ephemeral style intent, not just replenishing past buys. Nykaa has a gap in combining its strong editorial curation with underlying personalization tracking.
Moderate to Critical problem. Why: While users can still buy products without personalization, the friction directly impacts conversion and retention. In a highly competitive e-commerce landscape, an uncurated experience for a returning logged-in user is a massive missed revenue opportunity and a driver for platform switching.
Solving this problem creates meaningful value.
- Market size: Tens of millions of Nykaa Fashion active users.
- User willingness to adopt: High. Users naturally gravitate towards "For You" feeds when presented effectively (e.g., TikTok, Instagram, Sephora).
- Distribution difficulty: Low. The audience is already on the platform; this is an optimization of existing surfaces rather than building a new acquisition channel.
Core feature: A "For You" product shelf injected into the homepage and a re-ranked search results page based on a simplified formula: Historical Affinity (Brands/Categories bought) + Real-time In-session Intent (Last 3 clicks).
Intentionally excluded:
- Complex collaborative filtering (ML models).
- Cold-start 3-swipe module (too much front-end disruption for a V1 test).
- Computer Vision attribution tags.
What learning the experiment should generate: Does a basic rule-based recommendation shelf lift CTR and Add-to-Cart rates for logged-in users compared to the default editorial shelf?
- Technical risk: High latency for re-ranking search results could hurt UX more than the improved relevance helps.
- Market risk: Users might find the recommendations inaccurate if the rule-based engine is too simplistic, leading to "filter bubble" fatigue.
- Distribution risk: Low, as it is tested on existing traffic.
- Execution risk: Balancing the Category Team's manual boost levers with the personalization algorithm might lead to conflicts (e.g., pushing high-margin but irrelevant items).
Build