Amazon Sales Sentiment Analysis

code behind the report → Jupyter Lab Sentiment Analysis

---

Table of Contents

1. Background
2. Executive Summary
3. Dataset Overivew
4. Data Cleaning and Preprocessing
5. Sentiment Analysis Approach
6. Insights Deep-Dive
    6.1. Sentiment Overview
    6.2. Sentiment Score vs. Numerical Variables
    6.3. Highest and Lowest Sentiment Scores Across Product Categories
    6.4. Highest and Lowest Sentiment Scores Across Product Subcategories
           6.4.1. High Popularity Subcategories
           6.4.2. Low Popularity Subcategories
    6.5. Highest and Lowest Sentiment Scores Among Individual Products
           6.5.1. High Popularity Products
           6.5.2. Low Popularity Products
    6.6. Sentiment Trends in Popular vs. Niche Products
           6.6.1. Most Popular Products
           6.6.2. Least Popular Products
7. Recommendations
    7.1. Mid-to-High Review Volume (20k–200k), High-Sentiment Products
    7.2. Low-Review (<20k), High-Sentiment Items
    7.3. Stricter Quality Control & Product Improvement
    7.4. Implementation Notes

---

1. Project Background

[ back to contents ]

Amazon, the world’s largest online retailer, offers a vast selection of products across categories such as electronics, home goods, and office supplies. This sentiment analysis, based on January 2023 internal sales and review data, examines customer sentiment trends, identifies product quality concerns, and assesses sentiment correlations with price, discounts, and review volume. The findings guide ranking adjustments, quality control measures, and policy improvements to optimize marketplace efficiency.

2. Executive Summary

[ back to contents ]

This project analyzed Amazon sales data and customer sentiment across various product categories using a lexicon-based approach. The analyses revealed an average sentiment score of 0.813 (on a -1 to +1 scale) with 72% of products classified as positive, indicating a generally favorable buyer experience. A Pearson correlation analysis (r = 0.12) indicated that customer sentiment is weakly related to price, emphasizing the importance of product quality and user experience. Certain subcategories, such as battery chargers (electronics) and exhaust fans (home & kitchen), showed negative sentiment, highlighting the need for stricter quality control. Separating products by review volume (≥20k vs. <20k reviews) identified under-marketed but high-quality items, including webcams, power LAN adapters, and tripods, which could benefit from targeted promotions. To enhance customer satisfaction and marketplace efficiency, re-program search ranking algorithms to prioritize well-rated yet under-marketed products (e.g., Philips GC1905 Steam Iron and Mi 108 cm Full HD Android LED TV) while enforcing stricter quality measures for consistently low-rated items (e.g., ENVIE ECR-20 Battery Charger and Fire-Boltt Ninja Calling Smartwatch). Additionally, policy refinements should address recurring complaints in low-scoring subcategories, such as streaming clients and data dongles.

3. Dataset Overview

[ back to contents ]

The dataset consists of multiple product attributes, including customer reviews, ratings, and pricing details, summarized below:

Column Name	Description
product_id	Unique identifier for each product
product_name	Name of the product
category	Main category and subcategories
discounted_price	Discounted price of the product
actual_price	Actual price before discounts
discount_percentage	Discount percentage
rating	Average product rating (out of 5)
rating_count	Number of users who rated the product
about_product	Product description
user_id	Unique user identifier
user_name	Name of the user who left a review
review_id	Unique review identifier
review_title	Short review summary
review_content	Detailed review content
img_link	Image URL of the product
product_link	Official product page link

4. Data Cleaning & Preprocessing

[ back to contents ]

• Converted Price Data:
  • Removed currency symbols and commas to convert price data into numeric format.
• Category Truncation:
  • Inserted spaces and truncated long category strings from the category column for ease of reading.
    Before:
    Electronics|HomeAudio|MediaStreamingDevices|StreamingClients
    After:
    Electronics | Audio Devices ... | Streaming Clients
• Handling Missing and Duplicate Values:
  • Two products entries were removed due to missing rating count.
  • One product entry was removed due to non-numeric entry under rating.
  • 128 product entry duplicates were removed.
  • Initial Rows: 1,462 → After Cleaning: 1,334

5. Sentiment Analysis Approach

[ back to contents ]

Sentiment analysis was conducted using VADER (Valence Aware Dictionary and Sentiment Reasoner), a lexicon- and rule-based tool designed for informal text, making it particularly well-suited for analyzing product reviews and social media posts. Each product’s star rating and review text were used to calculate a Sentiment Score, ranging from -1 (most negative) to +1 (most positive). To enhance interpretability, products were also classified into Sentiment Categories: Positive, Mixed Negative, Neutral, or Mixed Positive, based on the distribution of their ratings and textual reviews. Higher sentiment scores indicate greater customer satisfaction, whereas negative scores suggest unfavorable opinions.

6. Insights Deep-Dive

[ back to contents ]

6.1. Sentiment Overview [↑]

• The average sentiment score was 0.813, indicating generally positive experiences.
• Positive sentiment was found in 72% (962 out of 1,334) of products.
• Only ~3% registered mixed-to-negative sentiments, and none were purely negative.

6.2. Sentiment Score vs. Numerical Variables [↑]

• A Pearson correlation analysis was conducted to examine the relationship between Sentiment Score and key numerical factors, including Actual Price, Discounted Price, Discount Percentage, Rating Count, and Rating.
• The results showed that all price-related factors had a weak or near-zero correlation (0 to 0.1) with sentiment, suggesting that customers prioritize product quality and overall experience over pricing considerations.

---

6.3. Highest and Lowest Sentiment Scores Across Product Categories [↑]

The main categories were ranked based on mean sentiment score using a bar chart, with colors representing the number of reviews.

• Highest-rated category:
  • Musical Instruments (0.996 ± 0.00276)
• Lowest-rated categories:
  • Computer & Accessories, Electronics, Office Products, and Home Kitchens show high score variability.
  • Despite lower average scores, some items still received extremely positive ratings.
  • Their large review volumes warrant further breakdown into subcategories.

---

6.4. Highest and Lowest Sentiment Scores Across Product Subcategories [↑]

Because subcategories vary significantly in review counts, they were analyzed separately:

1. High Review Counts (≥20k reviews):
   • Top 10 (highest sentiment scores)
   • Bottom 10 (lowest sentiment scores)
2. Low Review Counts (<20k reviews):
   • Top 10 (highest sentiment scores)
   • Bottom 10 (lowest sentiment scores)

In all cases, color bars represent review volume.

6.4.1. High Popularity Subcategories [↑]

• Top 10 subcategories display strong sentiment scores with low variability, suggesting consistent customer satisfaction.
• Highest performer: Bluetooth Adapters (Computer & Accessories) ranked first, highlighting reliability & popularity.
• Hidden opportunities:
  • Webcams, Memory, Power LAN Adapters, and Tripod Legs have high sentiment but moderate reviews, suggesting marketing potential.
• Improvement needed:
  • Battery Chargers (Electronics) and Exhaust Fans (Home & Kitchen) reflected negative sentiment, signaling quality gaps.

6.4.2. Low Popularity Subcategories [↑]

• Top 10 low-review subcategories:
  • Generally minimal differences in sentiment, all scoring relatively well.
  • These subcategories could benefit from better promotion.
• Bottom 10 low-review subcategories:
  • Streaming Clients (Electronics), Electric Grinders (Home & Kitchen), and Data Cards/Dongles (Computers) each scored below -0.5.
  • These weak sentiments suggest product quality issues or unmet expectations.

---

6.5. Highest and Lowest Sentiment Scores Among Individual Products [↑]

Products were categorized based on review count:

• High-review products (≥20k reviews)
• Low-review products (<20k reviews)
• Each segment’s top 10 and bottom 10 were examined.
• Color bars represent review count.

6.5.1. High Popularity Products [↑]

• Overall strong performance among the top 10 with minimal sentiment variation.
• Realme Wireless Buds (Yellow) excelled with a high sentiment score despite 200k+ reviews.
• Potential for growth:
  • Philips GC1905 Steam Iron (Blue)
  • Mi 108 cm Full HD Android LED TV 4C
  • Samsung Galaxy M33 5G (Mystique Green, Emerald Brown)
    These products have high sentiment but fewer reviews, indicating marketing expansion opportunities.

6.5.2. Low Popularity Products [↑]

• Products scoring below -0.5 warrant immediate review:
  • AmazonBasics ABS USB-A to Lightning Cable (3-Ft, White)
  • ENVIE ECR-20 AA & AAA Battery Charger
  • Dell MS116 USB Wired Optical Mouse
  • Fire-Boltt Ninja Calling Smartwatch
• These products require closer scrutiny, as negative sentiment may worsen with increased sales.

---

6.6. Sentiment Trends in Popular vs. Niche Products [↑]

Products were analyzed by popularity, using mean sentiment scores in bar charts.

6.6.1. Most Popular Products [↑]

• Top 3 most reviewed products:
  • Amazon Basics High-Speed HDMI Cable (426,973 reviews)
  • Boat Bassheads (363,713 reviews)
  • Redmi 9 Active (313,836 reviews)
• Despite massive review counts, these products maintain sentiment scores above 0.9, indicating high satisfaction despite large user bases.

6.6.2. Least Popular Products [↑]

• Each product in this group has under 8 reviews.
• Most products score above 0.6, implying moderate to positive experiences among the small number of reviewers.
• These products could benefit from increased promotion.

---

7. Recommendations

[ back to contents ]

7.1. Mid-to-High Review Volume (20k–200k), High-Sentiment Products [↑]

Subcategories to Consider

• Webcams
• Memory
• Power LAN Adapters
• Tripod Legs

Products to Consider

• Philips GC1905 Steam Iron (Blue)
• Mi 108 cm Full HD Android LED TV 4C
• Samsung Galaxy M33 5G (Mystique Green, Emerald Brown)

Why It Matters

These products already have a substantial review base (above 20k) and maintain strong sentiment, but they have yet to achieve “blockbuster” status with 200k+ reviews. Elevating them in search results and promotional campaigns can help tap into their latent potential, moving them closer to top-tier performance.

Key Actions

1. Targeted Promotions. Launch product- or category-specific ads, email campaigns, and homepage features to boost their visibility. Encourage cross-promotion where subcategories like Webcams or Memory are spotlighted alongside best-selling accessories.
2. Refined Search Placement. Adjust search algorithms to reward high-sentiment products in the 20k–200k review range, ensuring they appear prominently for relevant keywords. Monitor conversion and engagement metrics for these items, dynamically updating placements to maximize growth.

---

7.2. Low-Review (<20k), High-Sentiment Products [↑]

Subcategories to Consider

• Fountain Pens
• Tablets
• Espresso Machines
• Coffee Presses

Products to Consider

• Egate i9 Pro-Max
• American Micronic
• Camline Elegante Fountain Pen

Why It Matters

Though these items have fewer than 20k reviews, they exhibit high sentiment scores, indicating quality and customer satisfaction. With strategic marketing, they could grow substantially, bridging the gap between early positive feedback and broader market adoption.

Key Actions

1. Awareness Campaigns. Showcase these products in curated lists (e.g., “Top-Rated Essentials”) and utilize newsletters or social media posts to reach new audiences. Consider limited-time discounts to encourage trial and accelerate the review collection process.
2. Partnerships & Bundling. Bundle these under-reviewed, high-sentiment products with related high-traffic items (e.g., pairing Espresso Machines with popular coffee bean brands). Highlight subcategory benefits (e.g., Fountain Pens for professionals, Tablets for students) in cross-category promotions to draw in complementary buyers.

---

7.3. Stricter Quality Control & Product Improvement [↑]

Subcategories to Consider

• High-Popularity, Low-Sentiment (Immediate Action)
  • Battery Chargers (Electronics)
  • Exhaust Fans (Home & Kitchen)
• Low-Popularity, Low-Sentiment (Action Needed)
  • Streaming Clients (Electronics)
  • Electric Grinders (Home & Kitchen)
  • Data Cards/Dongles (Computers)

Products to Consider

• High-Popularity
  • Fire-bolt Earbuds
  • Dell MS116 Wired Mouse
  • ENVIE ECR-20 Charger Batteries
• Low-Popularity (Below –0.5 Sentiment)
  • AmazonBasics ABS USB-A to Lightning Cable (3-Ft, White)
  • ENVIE ECR-20 AA & AAA Battery Charger
  • Dell MS116 USB Wired Optical Mouse
  • Fire-Boltt Ninja Calling Smartwatch

Why It Matters

Despite strong sales volumes or established brand presence, these subcategories and products harbor chronic quality or performance issues. Ignoring negative feedback leads to higher return rates, dissatisfied customers, and potential brand damage.

Key Actions

1. Focused Product Audits. Conduct comprehensive technical evaluations, user feedback analyses, and manufacturer reviews, particularly for high-volume items with negative scores. Enforce stricter production standards and supplier accountability to mitigate recurring complaints.
2. Rapid Response & Remediation. Immediately address items scoring below –0.5 sentiment, prioritizing improvements in design, durability, or customer support options. Consider removing severely underperforming SKUs from listings until quality issues are resolved.

---

7.4 Implementation Notes [↑]

• Integration with Search Algorithms
  Incorporate both review-volume tiers and sentiment data into the ranking logic. Boost items that demonstrate strong potential for growth or improvement, while de-emphasizing those with persistently negative feedback until issues are resolved.

• Cross-Functional Collaboration
  Coordinate with marketing, product management, and supplier relations teams to ensure these recommendations translate into measurable actions—such as promotional campaigns, product design overhauls, or updated return policies.