The purpose of this project was to analyze Amazon Reviews to determine whether or not the paid review program Vine introduced a bias in the star rating of products. The Vine program either pays or sponsors individuals in exchange for a review on the product's Amazon page.
The scope of the analysis was limited to automotive products only. The dataset used can be imported from this Amazon directory under https://s3.amazonaws.com/amazon-reviews-pds/tsv/amazon_reviews_us_Automotive_v1_00.tsv.gz .
After the data was put through an ETL pipeline, Google Colab was used to turn it into usable dataframes which were in turn saved into a csv file. The resulting file was then further analyzed to answer the question central to this project: do paid Vine reviews create a bias in Amazon products star rating?
The script used to analyze the reviews yielded the following results:
- There were 82 reviews which WERE a part of the Vine program.
- This included 33 5-Stars reviews.
- 40.24% of the total were 5-Stars reviews.
- There were 24742 reviews which were NOT a part of the Vine program.
- This included 12807 5-Stars reviews
- 51.76% of the total were 5-Stars reviews.
Although there is a difference of 11.5% of 5-star ratings between regular and vine reviews, it is worth noting that Vine reviews only account for .03% of total reviews. The sample size being this much smaller may have skewed the results for the paid reviews ratings.
It seems, as far as this analysis is concerned, that Vine reviewers tend to give less 5-star ratings than unpaid reviewers.
Further analysis is needed to get a better grasp of whether or not the Vine program causes a rating bias. Options include the following:
- Test all categories rather than only the automotive reviews.
- Check for bias among all other ratings as well rather than only 5-star reviews.
- Look at differences between the amount of Vine reviews across categories.