🛒 Amazon Sales Performance & Customer Satisfaction Analysis

This project evaluates the transactional dynamics of Amazon sales data, focusing on the efficacy of discount strategies against customer satisfaction metrics (customer ratings). The analysis aims to dissect consumer purchasing behavior across various product categories to optimize profit margins and market retention.

🎯 Core Business Objectives & Focus Questions

1. Discount Optimization: Does an aggressive price-cutting strategy correlate positively with customer satisfaction?
2. Category Dynamics: Which product categories exhibit the highest elasticity towards promotional discounts?
3. Performance Anomalies: What core factors drive below-average ratings on heavily discounted items?

---

🛠️ Data Engineering & Preparation Rigor

Raw data processing was executed using Microsoft Excel, focusing on resolving structural inconsistencies and ensuring high data hygiene:

• Targeted Imputation & External Validation: Identified an anomalous string character "|" within the rating column for product_id: B08L12N5H1 (Eureka Forbes Vacuum Cleaner). Rather than dropping the row, external validation was conducted directly against the live Amazon India product page to accurately restore the real value to 4.0.
• Inconsistency Mitigation: Removed 2 rows containing permanent missing values in the rating_count column to secure the stability of statistical estimations.
• Data Type Transformation: Cleansed textual currency symbols from the discounted_price and actual_price columns, casting them into numeric formats suitable for quantitative processing.
• Category Granularization: Split composite string values using the "|" delimiter to cleanly extract the primary Main Category and Sub-Category.
• Feature Engineering (Rating Discretization): Created a new categorical performance metric (rating_score) mapped as follows:
  • < 2.0: Poor | 2.0 - 2.9: Below Average | 3.0 - 3.9: Average | 4.0 - 4.9: Good | 5.0: Excellent.

---

📊 Exploratory Data Analysis (EDA) & Key Findings

1. Macro Discount Profile

The baseline average discount across all product assortments stands at 47.64%. This underscores a heavy operational reliance on price-driven promotional campaigns to acquire volume.

2. The Discount vs. Satisfaction (Rating) Paradox

When non-discounted products (0% discount) were isolated from the experiment, a distinct downward trend in customer satisfaction emerged as discount percentages climbed (Negative Correlation).

• R-Squared ($0.0259$): Demonstrates that consumer satisfaction variance is heavily dominated by intrinsic product characteristics (build quality, expectations) rather than price cuts.
• P-Value ($< 0.0001$): Highly statistically significant. This confirms that aggressive discounting negatively impacts perceived value or signals "clearance product" low-tier quality, rather than driving customer delight.

3. Cross-Category Discount Distribution

The highest average promotional discounts are heavily concentrated within the Home Improvement (57.50%) and Computers & Accessories (53.92%) sectors. Conversely, Office Products maintain structural price integrity, recording the lowest baseline discount average (12.35%).

4. High-Ticket Item Cohort Performance

Premium category electronics, such as the Sony Bravia 65" 4K Ultra HD TV and the OnePlus 65" Smart Android TV, consistently retain the highest localized market prices despite undergoing deep nominal discounts.

5. Consumer Satisfaction Distribution Overview

Overall market sentiment resides in a healthy cluster, with 75.70% of products securing a Good rating score, while a minor fraction (0.41%) falls into the Below Average risk bracket.

---

🎛️ Tableau Interactive Dashboard

The complete analytical dashboard engineered for deep interactive exploration of these metrics is publicly hosted and accessible on Tableau:

➡️ Click Here to Open the Interactive Dashboard Workspace

---

💡 Strategic Business Recommendations

• Calibrate Extreme Discount Thresholds: Restrain the deployment of extreme discounts exceeding 50% in risk-vulnerable categories, as empirical evidence highlights an inverse relationship with customer perception and post-purchase satisfaction.
• Granularize Negative Feedback Loops: Initiate deep text/sentiment analysis specifically targeting products in the Below Average segment (0.41%) to isolate whether complaints stem from physical manufacturing defects or marketing mismatches.
• Value-Driven Premium Frameworks: Pivot high-ticket products (such as large screen TVs) away from pure price-cuts and towards Value-Driven Marketing bundles to protect premium brand equity and secure healthier operating margins.

👤 Author

Wira Dhana Putra

• Career Switcher → Moving purposefully into Data Science, Machine Learning, and AI Engineering.*
• Dataset Source: Kaggle Amazon Sales Dataset

📄 License

This project is licensed under the MIT License - see the LICENSE.md file for details.

---