Amazon_Vine_Analysis

Project Overview

The project will provide two deliverables for Amazon Product Review Datasets:

• Deliverable 1: ETL on Product Review Datasets provided by Amazon.
  • Amazon S3, Amazon RDS and pgAdmin will be used.
  • The data in pgAdmin will be seperated into the client's requested table sets
• Deliverable 2: Analyze the reviews to determine if there is any prositivity bias in reviews participating in Amazon's Vine Review Program
  • Technologies used: Google COLAB, PySpark

Additional Information:

• Amazon Vine Program: is a paid review program by Amazon to ensure quality reviews for products on their platform. It is an exclusive program only offered to users who meet Amazon's criterion, and the reviewers are also paid by Amazon. More information can be found here: https://www.amazon.com/gp/vine/help
• Dataset Source: https://s3.amazonaws.com/amazon-reviews-pds/tsv/index.txt

Deliverable 1: ETL Script

Amazon_Reviews_ETL.ipynb
The original data and columns were transformed and loaded into the following table (and columns)

Original Table	Customer Table	Product Table	Review_ID Table	Vine Table
marketplace	customer_id	product_id	review_id	review_id
customer_id	customer_count	product_title	customer_id	star_rating
review_id			product_id	helpful_votes
product_id			product_parent	total_votes
product_parent			review_date	vine
product_title			verified_purchase
product_category
star_rating
helpful_votes
total_votes
vine
verified_purchase
review_headline
review_body
review_date

Deliverable 2: Vine Review Analysis

• Dataset Used: https://s3.amazonaws.com/amazon-reviews-pds/tsv/amazon_reviews_us_Watches_v1_00.tsv.gz
• Link to script: Vine_Review_Analysis.ipynb

Key Results

Observations:

• The sample size of Vine (Paid) reviews is much smaller than non-Vine (Unpaid) reviews
• The ratio of 5 Star Reviews to Total Reviews per item is: 31.2% (Paid) and 51.8 (Unpaid) Interpretation:
• Definitive conclusions should not be made using this dataset as the sample size of 47 Paid Reviews is not sufficiently large
• However, from the above dataset, there is no positivity bias observed. In fact, bias looks to skew the other way.
• One possible explanation is that the Vine program eliminates bot/fake 5 star reviews undetected from Amazon's standard approach

Additional Recommendations

The client is recommended to invest additional resources to allow the analytics team to perform the same analysis on atleast a handful of more datasets with the main objective to capture more Vine (Paid) reviews.