Online retail business has become an essential of the 21st century. Alongside the benefits of convenience and accessibility, there is also a cumbersome amount of consumer data. One way online businesses have worked to maximize their sales is through marketing their product by reviews. Some companies participate in Vine programs, which incentivize consumers to leave review by providing them payment or discounted products.

The online retail giant, Amazon, has implemented this strategy. Using PySpark, we will look at the Amazon customer reviews of electronics (provided by this dataset). By wrangling the values in these data, we will assess what influence Vine programs have on the positiveness of reviews, if any at all.

The following table represents the counts acheived through extracting data from the source. (See Vine_Review_Analysis.ipynb for source code.)

There were 1029 Vine reviews (incentive=Paid) and 46593 non-Vine reviews (incentive=Unpaid) .

There were 440 five-star Vine reviews (incentive=Paid) and 21710 five-star non-Vine reviews (incentive=Unpaid) .

Of the 1029 Vine reviews, 42.8% were five-star reviews. Of the 46593 non-Vine reviews, 46.6% were five-star reviews.

The data indicate that there is no positivity bias presented by the Vine reviewers. In fact, the non-Vine reviewers provided proportionally more five-star reviews.

To extend this analysis, it would be interesting to see how these numbers change if we were to keep all reviews regardless of how many helpful_votes they received (in this dataframe, these were filtered to only include reviews with more than 20 helpful_votes, and then only those reviews that exceeded 50% helpful_votes over total_votes).

Additionally, it would be interesting to see if there are correlations against Vine program reviews, ratings, etc… against verified_purchase. This would provide better veracity of those reviews.