page contents



Print this Post

Research: Underdog Businesses Are More Likely to Post Fake Yelp Reviews

The rise of online review platforms like Yelp has empowered consumers by reducing the informational asymmetry surrounding unfamiliar products and services. But some businesses face strong incentives to post fake reviews, which compromises the trustworthiness of such review sites.

How prevalent are such fake reviews, and what firms are mostly likely to post them? New research reveals that independent businesses, those without many existing reviews on Yelp, and those that face intense competition are more likely engage in review fraud on Yelp.

I reached out to authors Michael Luca, Professor at Harvard Business School, and Georgios Zervas, Assistant Professor of Marketing at Boston University School of Management, to discuss their findings. Here’s an edited transcript of our conversation.

How did the two of you become interested in this line of research?

Michael Luca: We’ve both done extensive work on consumer review websites in the past, and I’ve seen in prior research that Yelp reviews are incredibly influential on sales. Because of this influence and because these types of platforms have proliferated over the last decade – think of Yelp, TripAdvisor, Angie’s List, and scores of others – it’s important to understand the pros and cons of this type of system relative to other information sources.

Georgios Zervas: I was working on reviews as well but from a different perspective. I was trying to see what actually shapes reviews, so I was looking at online promotions like Groupon and LivingSocial and trying to explain why these coupons tended to result in negative reviews on Yelp. So we had both done work on reviews, and we joined forces to work on fake reviews.

According to your research, what leads businesses to post fake reviews on Yelp?

Georgios: The incentives are economic. The system is set up in such a way that businesses can benefit a lot from soliciting fake reviews. So businesses will respond to incentives such as a bad recent reputation, or having few reviews, or generally being unknown.

You’ve shown that some new businesses post fake positive reviews when they have very few existing reviews. Conversely, established businesses (including chains) are less likely to post fake reviews. Is it possible that high-quality firms might use fake reviews as a form of free but fraudulent advertising, from which they “graduate” as their reputation builds?

Georgios: Yes, I think that makes a lot of sense. For example, restaurants may be very resource-constrained, so when a new restaurant is starting out, they may not have enough resources for advertising, and of course, fake reviews are seen by a lot of people to have a very low production cost. Once these restaurants become established, potentially they’ll have more resources to invest in advertising so it’s very logical that they would graduate to more legitimate forms of advertising.

Besides misleading consumers, are there any other economic costs or inefficiencies that fake reviews cause?

Michael: Just the mere presence of fake reviews has led to an arms race in the review industry. These review sites dump resources into solving the fake review problem by making it more difficult for someone to leave a review, or by forcing a reviewer to jump through more hoops to make sure he or she is real, or by filtering some reviews off of the site, as Yelp does. Every barrier that you construct on these types of platforms will help reduce the number of fake reviews, but also actually reduces the number of real reviews on the site.

Michael, you mentioned earlier that Yelp reviews can significantly affect sales. How can we quantify that effect?

Michael: The way to think about returns to reviews is in the value of the rating. Say the user goes onto Yelp and sees a business with 3.5 stars and 10 reviews next to it. And the estimates we saw in prior work is that a one-star increase in rating leads to more than a 5% increase in sales for independent businesses, which is a really huge effect. If you think about your ability to influence this at the margin, especially when you don’t have many reviews, you have a very high-powered incentive to leave fake reviews.

In future work, what are some other independent variables about a business (like social media engagement, which you’ve mentioned) that you want to correlate with propensity to commit Yelp fraud?

Michael: One thing that we think would be exciting is the relationship between fraud and advertising. We see that people who are committing fraud have just had a negative reputational shock, are facing intense competition, or are early on in their business life cycle. On the other hand, we have work that suggests that firms using advertising have established reputations, have had positive reputational shocks. So our plan for our follow-up paper is actually to look through and think about the relationship between review fraud and advertising decisions. The question is, do firms face a substitution decision between advertising and review fraud?

How can a content aggregator like Yelp use this paper to improve its ability to detect fake reviews?

Georgios: The detection algorithms for sites like Yelp work by going through a set of reviews and carefully classifying them into real and fake reviews. Then they try to figure out some distinguishing characteristics of fake reviews. For example, fake reviews might come from users that have only written one or two reviews on Yelp ever. Then they use that indicator as a positive signal of a review being suspicious for new reviews that arrive on the site. We suggest that all these features we’ve identified – a business being new, or a business having recently received bad reviews, or a business facing increased competition – could be incorporated with those prior features into existing algorithms to predict whether a review is fake or not.

How might a small business with a Yelp presence use this research to improve its own sales?

Michael: One simple example is to engage with the system in an ethical way: claim your business on Yelp, make sure the information and hours are correct. This kind of persona can generate more legitimate reviews without gaming the system. Small businesses can also understand how exactly the Yelp system works. If you ask 10 small businesses how Yelp decides to aggregate reviews, you might get 10 different answers. Spending a couple of hours to understand the system and what exactly is being displayed on the other end could reduce the anxiety of somebody who otherwise views Yelp’s algorithm as a black box.

What should customers who use Yelp know?

Michael: The customer should realize that some reviews might be illegitimate, and our work gives a sense of the situations in which you’re most likely to see a fraudulent review.

Georgios: Most users may just look at the rating itself. Our work tells the consumer that he actually needs to read the reviews, inspect the business, ask about their advertising, ask whether it’s a chain, see if it’s in a highly competitive area, and then make up his mind.

To wrap up, many of your empirical results are consistent with and thus powerfully supported by theoretical predictions. Did any of your results instead surprise you by going against prediction?

Georgios: One thing that surprised us both was our very first finding – that approximately 16% of all reviews are filtered by Yelp. Nobody had measured that before. So while you could see the number of filtered reviews for an individual business, nobody had done an extensive site-wide measurement of how many reviews were suspicious.

Michael: We’ve found that the “bad apple theory” isn’t quite valid. In our research, we both talked to a lot of small businesses, and we don’t have the sense that there are these small businesses going around trying to cheat the system. Something we didn’t see fully beforehand is that there’s simply a lot of pressure for a small business starting out without an established reputation to go and do something that doesn’t necessarily seem unethical at the time. But taking a step back, these businesses will often recognize the ethical implications of what they’re trying to do.


About the author

Mystic Maggie

All of the Mystic Maggie Posts are RSS Reader Feeds from around the web. All copyright remains with the original publisher. No copyright is claimed or intended. Where supplied, links back to the original article are included.

Permanent link to this article:

Powered by WordPress Lab
Powered by Yahoo! Answers