Sustainability Dashboard for the product seller or owner
Eco Rank based Product Ranking

EcoRank: LLM-Powered Sustainable Product Reranking

🧠 What Inspired Us

While Amazon already offers sustainability certifications and filters like “Climate Pledge Friendly,” these features remain static and binary, lacking granular, ranked insights for users. There is currently no dynamic sorting that lets shoppers view products based on a quantitative sustainability score across multiple dimensions.

As environmentally conscious individuals, we saw an opportunity to go beyond simple filtering. We were inspired by Amazon’s own Climate Pledge and the broader shift toward carbon-aware, responsible consumerism. Our goal was to reimagine product discovery by making eco-friendliness a ranked, explainable, and user-controllable dimension—just like price or delivery speed.

EcoRank can fill this gap—providing consumers with actionable, AI-driven sustainability scores that help them shop smarter and greener at scale.

Key Idea & Innovation

EcoRank is an AI-powered reranking system for e-commerce search that uses Large Language Models (LLMs) to compute a product's Sustainability Score. This score is based on material usage, packaging, shipping emissions, and lifecycle factors—then used to reorder search results with an “EcoRank” toggle for customers.

Core Innovations:

✅ LLM-Driven EcoScore for every product
✅ Real-time Reranking based on sustainability
✅ AI-generated impact explanations for transparency
✅ Seller Dashboard with improvement insights

Technical Architecture

1. LLM Scoring Engine

Model: Hugging Face’s Qwen-1.5B-Instruct
Input: Product title, materials, packaging, shipping origin, buyer location
Output: Score (0–100) + 1–2 line explanation

2. Backend Pipeline

Extract product metadata from structured/unstructured data
Send prompt to LLM and receive eco analysis
Cache results for fast search reranking

3. Frontend Interface

Amazon-style UI with toggle between "Best Match" and "EcoRank"
Displays EcoScore and explanation in product card
Dashboard view for sellers with sustainability suggestions

📊 EcoScore Calculation Methodology

Weighted Formula:

Should be tailored to established sustainability metrics leveraging the following factors EcoScore = (Material_Score × 0.35) + (Packaging_Score × 0.25) + (Shipping_Score × 0.20) + (Lifecycle_Score × 0.20)

Scoring Factors:

An example of scoring could be like following.

Material Impact (35%)
- High: Bamboo, Recycled plastic → 85–100
- Medium: Wood, Glass → 50–70
- Low: Virgin plastic, PVC → 10–40
Packaging Impact (25%)
- High: Plastic-free, compostable
- Medium: Recyclable cardboard
- Low: Styrofoam, excess plastic
Shipping Impact (20%)
- Based on distance between origin and buyer
- Local (<500mi): 90+
- Regional (500–1500mi): 50–80
- Long-haul (>1500mi): 10–50
Lifecycle Impact (20%)
- Based on durability, reusability, end-of-life disposal
- Inferred from product type, reviews, and materials

💼 Business Impact

🛍️ Improved Consumer Trust: Eco-conscious users can shop confidently
📈 Revenue Uplift for Green Sellers: High EcoScores can boost conversion
🧠 New Discovery Layer: Adds sustainability as a sorting/filtering dimension
💸 Data Monetization: Offer EcoRank insights as a premium feature or API
🛠️ Supports ESG Goals: Helps Amazon align with climate commitments

🌍 Sustainability & Social Impact

🌱 Carbon Reduction: Prioritizes local, low-emission shipping
♻️ Waste Reduction: Encourages compostable and minimal packaging
🧪 Sustainable Materials: Promotes adoption of eco-friendly materials
💡 Public Awareness: Increases consumer understanding of impact
🔁 Circular Economy: Supports reuse and recyclability through lifecycle data

🌟 Key Features

✅ EcoScore (0–100) with natural language explanation
✅ LLM-Driven Analysis using Qwen-1.5B-Instruct
✅ Dynamic Product Reranking toggle in UI
✅ Seller Dashboard with actionable improvement suggestions
✅ Multi-factor Scoring: Material, Packaging, Shipping, Lifecycle

📈 Key Metrics & KPIs

System Metrics

⚡ EcoScore Latency: <100ms (cached), <15s (live scoring)
🎯 Model Accuracy: Cross-validated with known eco-certifications
📊 Toggle Usage: % of users switching to EcoRank view

Business Metrics

📈 Sales Lift: Increase in high EcoScore product sales
🧮 Seller Adoption: Use of dashboard insights
💰 Revenue Impact: Premium pricing potential for top-ranked products

Sustainability Metrics

🌍 Estimated CO₂ Reduction from local shipping prioritization
🔁 Material Shift: Increase in sustainable inputs
📦 Packaging Improvement: Drop in plastic-based packaging

📚 What We Learned

🧠 Prompt Engineering: Careful formatting drives better LLM output
🏗️ Score Design: Scoring needs balance across diverse product types
🧾 LLMs as Scorers: Great for soft metrics like sustainability when structured inputs are lacking
🤝 Consumer-Seller Dynamic: Clear incentives help drive greener behaviors