New Research Uses Text Mining to Reveal More in Reviews
When it comes to reading business reviews online, the five-star rating is king. But new research from Maryland Smith says there is much more to learn from reviews than just the star rating.
For businesses like restaurants, hotels and spas, most consumers want to know primarily about the service quality they will receive, but measuring service quality in those spaces has been tricky for a long time, says Maryland Smith’s Anandasivam Gopal.
“There’s always been this gap in the field about measuring service quality in terms of being able to scale up, to find a way to measure services in a way that’s not intrusive and compare service quality measures across competitors,” says Gopal, the Dean's Professor of Information Systems at the University of Maryland’s Robert H. Smith School of Business.
The research, which was created as part of a dissertation from Jorge Mejia’s, PhD ’16, and subsequently published in Manufacturing and Service Operations Management, explores using text-mining algorithms to extract information from blocks of text in online reviews.
In the case of restaurants, reviews included measures of satisfaction with respect to the food, wait staff and ambiance, Gopal says. To verify the results, the researchers compared their data with Zagat ratings from the past 10 years.
“We knew we needed to be able to pull out the right service quality measures using text mining and also verify that the new measures correlated with traditional measures,” says Gopal. “But the third important aspect we needed to show was that the new measures we identified are associated with an economic outcome that tends to be correlated with service quality.”
That economic outcome, Gopal says, based on the argument that restaurants who did badly by service quality metrics displayed a higher probability of going out of business. Extracting these service quality metrics can also help businesses better understand their standing among similar businesses in the same vicinity, industry and price point, he says.
“There’s a lot of information contained in these online reviews that, when appropriately used, can give us a way to measure service quality while overcoming the traditional measurement problems like having to run obtrusive surveys and scaling issues,” says Gopal.
Gopal says the research has garnered attention from two interesting sources. The first, he says, is from social media content aggregation services seeking to provide businesses with more information about what customers are saying about them online.
The second is from real estate agents who believe the research could offer insights into potential location openings months before they occur. That was more of an unintended consequence of the research since the point of the paper was not specifically to cater to the real estate sector, Gopal says, but it also serves as validation of the research’s merits.
“While we have established a correlation between low service quality scores and business closures, we didn’t intend for the research to be interpreted that way,” says Gopal.
“But what that also says is that our research can be really useful in the hands of consumers.”
Read the full research, “Service Quality Using Text Mining: Measurement and Consequences,” in Manufacturing and Service Operations Management.