• Data Source:
  • Vine Review csv
  • Sample Data - Amazon Video Game Reviews
• Software: AWS, PySpark, PostgresSQL, Google Colab, Python, Pandas

The Amazon Vine program is a service that allows manufacturers and publishers to receive reviews for their products. Companies like SellBy pay a small fee to Amazon and provide products to Amazon Vine members, who are then required to publish a review. I am tasked to determine if there is any bias towards reviews written as part of the Vine program (paid vs. unpaid reviews).

Steps Taken:

1. Using PySpark on Google Colabs, I performed the ETL process on a big data set of over 170,000 rows from sample data about Amazon Product Reviews.
2. Afterward, I created an AWS relational database instance and transformed the data to be loaded into PostgreSQL (SQL). After loading the data into PgAdmin, I exported the Vine Review Table as a CSV file.
3. Finally, I load the data into Python to create Dataframes using Pandas. Then I analyzed the data to determine if there was bias in paid reviews versus unpaid reviews.

Analysis-JupyterNotebook

• Amazon Video Game
• Reviews with 20+ Votes
• 50% or higher on helpful votes

Total Number of Reviews: 40565

• Paid Views: 94
• Unpaid Views: 40471

Total Number of 5-Star Reviews:

• Paid 5-Star Reviews: 48
• Unpaid 5-Star Reviews: 15663

Percentage Of 5-Star Reviews:

• Paid Reviews: 51%
• Unpaid Reviews: 39%

Average 5-Star Rating:

• Paid Reviews: 4.2
• Unpaid Reviews: 3.35

Overall, there appears to be a slight bias with Video Game Paid Amazon Reviews based on the overall percentage of 5 Star Reviews.

• Paid Reviews: 51%
• Unpaid Reviews: 39%

With that said, that same bias could be by looking at Average 5-Star Rating:

• Paid Reviews: 4.2
• Unpaid Reviews: 3.35 However, the 51% ratio is based on a relatively small share of the overall reviews (0.2%).

Since the overall share of paid reviews is small, I would need to conduct further testing among various Amazon product categories to confirm this bias. Therefore, I would apply the same analysis across 10-20 categories. If the results are consistent across the different Amazon product categories, I feel that paid reviews are biased.