Driving Seller Efficiency and Revenue in Nykaa Seller Portal
Predictive Ordering automatically recommends what SKUs to restock, when, and in what quantity by analyzing sales trends and demand patterns, helping sellers reduce stockouts and protect revenue.
​​
Defined and delivered seller-portal features by synthesizing user insights into structured requirements and user stories, and prioritizing the roadmap to improve engagement by 18%.
​​
Confidentiality Note: Due to organizational data privacy and security requirements, actual screenshots are not publicly shareable. The visuals presented here are illustrative samples reflecting the implemented functionality.
The Problem: Inventory Inefficiencies Resulting in Lost Sales and Low Seller Engagement
Nykaa’s marketplace sellers were facing frequent inventory stockouts and overstocking, leading to:
-
Lost sales due to unavailable products
-
Increased operational costs from excess inventory
-
Heavy reliance on manual merchandising and guesswork
-
Low trust in the seller portal as a decision-making tool
From Nykaa’s perspective, this directly impacted GMV, seller engagement, and platform reliability.
I worked as a Product Manager Associate within the seller platform team, owning the problem from discovery to launch.
​
​Identifying Seller Pain Points & Understanding Seller Needs
​
​
​
​
​
​
​
​
​​​​​
​
​
​Defining the Product
​​
Product Vision
​​​
Enable sellers to proactively manage inventory through predictive reordering by auto-generating restocking recommendations based on sales patterns and seasonality trends.
​​
Prioritization & Roadmap
​
-
Prioritized problems based on revenue impact, feasibility, and seller effort saved
-
Built a lean MVP focused on predictive reordering
-
Planned iterative enhancements post-MVP to validate market opportunity and scale adoption
​
​
​
​
​
MVP Scope
​​
-
Auto-generated weekly restocking recommendations using:
-
Historical sales data
-
Seasonality trends
-
Rule-based thresholds
-
​
​
Product Requirements & Execution
​​
Product Requirements Document (PRD)
​
-
Authored PRDs detailing:
-
User stories and acceptance criteria
-
Business logic for predictive alerts
-
Edge cases like seasonality spikes
-
-
Worked closely with:
-
Engineers (feasibility, APIs, data pipelines)
-
Designers (simple, glanceable UI)
-
Seller ops teams (workflow alignment)
-
Marketing teams (seller communication strategy, feature positioning, and launch enablement)
-
-
Ensured cross-functional alignment before development sign-off
​​​
Success Metrics
​​
-
Reduction in stockouts
-
Increase in revenue
-
Increase in active seller engagement
​
A/B Testing
​
-
Ran experiments by exposing a test cohort of sellers to predictive restocking recommendations while maintaining a control group on the existing workflow
​
-
Compared engagement, alert interaction, and restocking behavior between test and control cohorts to validate effectiveness
-
Used seller feedback and usage data to fine-tune alert thresholds and recommendation logic prior to full rollout
Launch
​
-
Phased rollout to high-engagement sellers to drive early adoption and validate impact on stockouts, revenue, and engagement
-
Partnered with Seller Ops and Marketing to support seller communication, onboarding, and initial adoption during rollout
​
My Role
Impact & Metrics
30% reduction in projected SKU-level stockouts
Driven by timely, data-backed restocking recommendations
10% GMV uplift for participating sellers
From recovered sales due to improved product availability
18% increase in active seller engagement
Reflected in higher weekly usage and interaction with restocking recommendations.
Future Iterations
2
Phase
Dashboard & Recommendation Explainability
-
Introduce a consolidated dashboard that summarizes restocking recommendations beyond SKU level (e.g., category-level and portfolio-level views), enabling sellers to assess inventory health holistically
-
Provide clear explainability for each recommendation by surfacing underlying drivers such as sales trends, seasonality signals, and demand patterns, increasing seller trust and confidence in automated decisions
3
Phase
Seller-Specific Lead Time Buffering
-
Enhance recommendation accuracy by incorporating seller-specific lead times, ensuring restocking alerts are triggered early enough to prevent stockouts despite vendor delays.
Key Learnings
-
MVPs are most effective when they reduce cognitive load, not just add data
Turning inventory complexity into clear, SKU-level actions drove faster adoption than exposing raw metrics.
​​
-
Seller trust increases when systems provide explainability, not just recommendations
Adoption improved when sellers understood the signals behind restocking suggestions.
​​
-
Behavioral metrics matter more than feature usage metrics
Measuring stockout reduction and reorder follow-through was more meaningful than clicks or views.