Decomposing TripAdvisor: Detecting Potentially Fraudulent Hotel Reviews in the Era of Big Data

The impact of customer reviews on user purchase decisions has been well documented. For example, a one-star increase in a restaurant's Yelp rating can lead to a 5 to 9 percent increase in revenue. Unfortunately, this has motivated some businesses in review-dependent industries to falsify reviews. In the era of big data, analytical methods have made detection of these false reviews easier. We perform a longitudinal study of 2.65 million hotel reviews made by nearly 320,000 reviewers on TripAdvisor (which does not verify customers stayed at the property they reviewed) and compare them to 2.93 million reviews on other two other booking platforms, Agoda and Booking.com (which verify its reviewers stayed at least one night at the property they reviewed). We analyze the language used, the patterns of reviewer activity, and the change in hotel's reputation score over time across more than 5.5 million reviews. We find the word frequency between the two types of websites and the patterns of reviewer activity differ considerably, even though the relative ranking of hotel reputation scores across review platforms are similar.