An increasing number of online users now examine review websites before buying products, booking hotel rooms and planning for other events, yet a recent MIT Technology Review found that upwards of 72% of all online reviews are fake.
Researchers developed a mathematical equation that allows them to examine the natural J-shaped curve that occurs among reviews.
As the team explains, reviews tend to offer clustered results, for example there will be many one-star reviews and only a few two, three and four star reviews than clusters groups of five-star reviews. Those clusters eventually form a J-shaped curve.
By examining reviews in which a person only leaves 1-2 reviews for a website researchers can determine if their review is outside of the statistical normality of reviews for the goods or service and then eliminate or keep that review. If the one-time review clusters differently than the multiple reviewers it is possibly a fake review.
The mathematical equation also flagged sudden clusters of reviews which are often used as part of positive marketing spin campaigns.
Using earlier studies and their own research the team at MIT found that 72% of online reviews were staged.
By using the new mathematical equation and pairing it with textual analysis websites such as TripAdvisor and Hotels.com could more accurately determine the validity of reviews and then isolate spam. While those sites will likely not remove reviews because of the potential removal of real reviews that fit outside of normative results, they could potentially create a new type of scoring system to better rank reviews. In any case determining which reviews are real still falls at the hands of online readers.