Amazon Vine is an invitation-only program where reviewers receive free items in exchange for product reviews. Vine reviews have a disclaimer on the Amazon website for transparency. This project aims to confirm there is no positive product review bais for these types of paid reviews. The analysis was performed in Google Colaboratory with PySpark and output as a ipynb.

This dataset of Amazon reviews in the Books product category. The schema includes:

• marketplace
• customer id
• review id
• product id
• book title
• star rating
• helpful votes
• total votes
• vine designation
• review inforamtion

To clean the dataset, I first filtered the data to include items where the total_votes >= 20 votes, and the helpful votes to total votes ratio was >= 50%.

Then I created two new dataframes that filter for Vine votes, to have two separate datasets for paid reviews and unpaid reviews. A simple count function of each dataframe shows the number of vine reviews (paid) and non-vine reviews (unpaid).

Filtering each dataset by the "star_rating" column, reveals the number of 5 star reviews for each group.

Comparing the percentage of 5-star reviews as a part of the whole gives us a better idea how many reveiws are 5 star across both groups (even with smaller population sizes).

There is no strong positivity bias for reveiws in the Vine program in the books category. 40.52% of Vine reviews were 5 stars, and 45.71% of unpaid reviews were rated 5 stars. One additional analysis to further confirm would be to compare the rating of the other star reviews to see if a similiar distribution of data or any skewing of the data.