ML-Driven Recycling & Rewards Concept

Kohl’s × Sephora × PACT Case Study

Project Context

Role: Conceptual System Designer (Case Competition Entry)
Objective: Propose a technology-backed solution to increase sustainability engagement in retail.

View the Full Group 3 Presentation (PDF)

The Challenge

Kohl’s and Sephora have a strong partnership, but bridging the gap between physical sustainability (recycling) and digital loyalty (Kohl’s Rewards) remains a manual process. The goal was to design a system that automates this bridge to drive foot traffic and environmental impact.

The Proposed Solution: “The Smart-Pact Bin”

The core concept is an AI-assisted recycling kiosk that identifies beauty product packaging and rewards the user instantly.

Conceptual System Architecture

The system would require a three-tier integration:

1. Hardware Layer: IoT-enabled bins with weight sensors and high-resolution cameras.
2. Processing Layer: A Computer Vision (CV) model to identify Sephora-brand packaging vs. non-recyclable waste.
3. Incentive Layer: Integration with the Kohl’s Rewards API to credit accounts in real-time.

Role of AI & Machine Learning

In this conceptual model, AI is the “trust layer” that removes the need for store associates to manually verify recycled items.

1. Object Recognition

Using a pre-trained image classification model, the bin would verify if the deposited item matches Sephora’s accepted materials (e.g., plastic bottles, glass jars, or tubes).

2. Weight-to-Point Logic

By combining visual data with a weight sensor, the system would estimate the mass of the material and apply a conversion formula:

• Formula Concept: $Points = (Weight_{material} \times Material_{value}) + Loyalty_{bonus}$

Feasibility & Cost Analysis

A major part of the study involved high-level financial modeling for a pilot program:

• Total Estimated Pilot Cost: ~$862,500.
• Hardware (Bins/Sensors): Primary cost driver.
• Maintenance & Logic: Assumed API integration costs and data handling fees.

Key Takeaways

This project allowed me to practice Technical Product Thinking—the ability to look at a business problem and break it down into modular technical components.

• System Design: Thinking about how IoT hardware interacts with cloud APIs.
• Data Integrity: Addressing how to prevent “reward fraud” using sensor data.
• User Experience: Designing a frictionless flow for the non-technical consumer.

Disclaimer

This project was developed for a case competition. It represents an exploration of technical feasibility and business strategy. All cost estimates and technical architectures are theoretical.This project is a conceptual proposal. No physical system was built, and no code was deployed. The following documentation outlines the high-level system architecture and the logic behind the proposed AI integration.

• 2025 3
• 2024 4
• 2023 2
• 2022 1