How Local Sites Can Win the War on Fake Reviews | Street Fight

Commentary

How Local Sites Can Win the War on Fake Reviews

0 Comments 30 August 2013 by

fakeIt’s well known that online reviews are one of the most important drivers of consumer decisions to buy or not buy. It’s no wonder that some businesses managers and marketers go to great lengths to secure good reviews, even if it means creating fake reviews and posting them to popular review sites. This week Edmunds, the auto and car dealer review site, announced that it had settled a lawsuit it had filed against Humankind, a firm accused of submitting fake reviews for a fee to about a dozen review sites on behalf of businesses. Humankind was accused of creating fake user accounts and attempting to submit glowing reviews of their clients’ car dealerships. The settlement terms state that Humankind:

  • Submit to a permanent injunction agreeing not to register users on Edmunds.com, not to submit reviews, not to participate in Edmunds.com’s online community, and not to breach the site’s membership agreement.
  • Provide Edmunds.com with information concerning all member accounts they registered and all reviews they submitted to Edmunds.com.
  • Pay an undisclosed sum toward Edmunds.com’s legal fees.

According to Aaron Lewis at Edmunds, one reason they caught the infraction was because the company uses humans to manually check each review. While this is a good way to increase the trust factor of reviews and can work well in conjunction with other methods it does come with tradeoffs and isn’t feasible for every service. No system is 100% full proof (including those of market leaders like Yelp & Google) and sites need to continuously invest in ways to improve their review filters as they identify new trends and analytics related to fake reviews.  Here are some other methods for your site or app that will catch a high percentage of fake reviews.

  • Bulletproof Your Terms & Conditions, Terms of Use, and Privacy Policy to include reviews on your site — clearly spell out what you expect from users, 3rd parties, and others. Reading the Edmunds Membership Agreement they did a good job of setting up their legal team with the tools needed in court. Having a written policy against fraudulent reviews will also limit a company’s liability. Hire an attorney that is experienced in Terms of Use, digital copyright, and understands the web/social/mobile space.
  • Captcha – Many underhanded operations that try and game review sites tend to follow the path of least resistance to save cost and time. In most cases this means writing reviews on a listing then using an automated system to create accounts and post the review. Having a captcha is one way to stop those using automation. It does come at a small cost where some valid users might drop off and not leave the review — surprisingly AB testing captcha with an explanation to the user of why, resulted in a smaller percentage of drop off of valid reviews than we were ready to accept as a tradeoff on Judy’s Book.
  • IP-capturing and using the IP of the user is a great tool. Fake reviewers will tend to try and submit multiple reviews from the same place, especially if it’s a firm.  You can even cross reference the IPs physical address with the address of the place that the review is being submitted on.  Realize with mobile it’s a valid case for an actual customer to write a review from the place but if it happens more than normal, your filter should raise a flag to investigate more.  Looking at the Geolocation of the IP, you should be suspicious of reviews from international locations for a site with only US listings.  Especially from India, Pakistan, Bulgaria, Vietnam, The Philippines, where US based firms tend to outsource review submission.
  • Industry — Not all categories are equal and we’ve found that certain industries seem to submit more fraud.  At times the tenacity of fake users from the following industries has caused more than 1 of our developers to lose a night of sleep and we’ve considered turning off whole categories of listings and removing from our site, but there were enough honest businesses that we decide to keep them.  Scrutinize Locksmiths, Garage Door Repair, Towing Services, and Moving Companies.  Even with the recent example with Auto Dealerships hiring Humankind I don’t put auto dealers in this bucket.
  • User behavior (ads image to profile, multiple reviews of different categories, reviews on places in same city as IP, login with Facebook).  This is one of the most effective signals in not only identifying reviews not to post but having your fraud filter periodically look at reviews already on your site and flag for investigation.  In the last year, 99% of reviews submitted on our site via user signing in/up with Facebook were deemed valid after an internal audit.  Facebook has gotten good at identifying fake accounts and purging them, for startups with limited resources this can be a good option.  Also for sites that span multiple categories the typical valid user tends to review places that span those categories.  Many marketing firms target clients in a certain vertical and a reliable signal of fakes can be a user who’s only reviewed Spas, Lawyers, Doctors.  Generally users who write multiple reviews tend to be more trustworthy and it’s completely valid for a user to review many eateries, but how often does the average person use a tow truck?  It’s valid for a user to review many hotels in different cities but how often have you used a plumber or electrician in a city you don’t live in? Get the hint?  Users who take time to add an image and more details to their profile tend to garner higher trust. On a 5 star rating scale most of the fake reviews will be a 5 star, some 1 star, and a very small amount 2-4 stars.
  • Text and Sentiment Analysis — It’s also possible to have your system view the text in the review(s) and using text from other known fake reviews help flag fakes.  At a minimum you should flag addresses and contact information inserted into reviews, multiple mentions of the place name, marketing speak, or a review that sounds like a press release.  Real reviews tend to mention the good, bad, service, and overall value.  An advanced system can periodically search your site for strings of words that appear in multiple reviews to flag for a human to investigate.  Building your own text or sentiment analyst tool is not a simple endeavor, I recommend using a service like Bitext.

For fake reviews that do get on your site, one way to identify them for removal is to harness your user community. Consumers have become good at spotting fakes; the key is to make it really easy and simple for users to report suspicious activity, and reward with praise for doing so. Most of the top review sites have a flag or report link next to every review. To retain user trust and limit liability, I think it’s a good practice to stop displaying the review and put it in a queue for a human to investigate upon the user reporting it. If the site admin determines it’s valid, it can be re-activated.

Another is to use them to deter other business owners and marketers from posting fakes. Yelp does this by shaming the business by posting a consumer alert on the places Yelp profile.  I like this but I don’t recommend it for a startup that doesn’t have the legal resources to back it up.

Edmunds is providing information to other review sites (that Humankind could have breached) to help them determine if they need to take action to protect their consumers. One way to win the war for consumers would be for brands to partner on strategies and share best practices.

Ali AlamiAli Alami is the Interim CEO of Judy’s Book and KidScore, local search and review sites that connect families with local places. Follow him on twitter @alialami.

Nov. 4th in NYC: Local in the City!
Click here to SAVE $500 thru Aug. 1st.

Newsletter

Get hyperlocal industry headlines in your inbox every morning. Subscribe to the Street Fight Daily newsletter.

Follow Us

Get the latest Street Fight news, information and analysis via Twitter and Facebook.

The Commerce Graph

The “Commerce Graph” is a new framework we have developed to think about the future of physical exchange. The model offers an alternative to the dominant narrative about the commerce landscape that frames digital networks as an adversary of physical exchange.

The $20 Billion Mobile Marketing Opportunity

Strategies and insights into the landscape of targeting options and how they deliver foot traffic and sales for SMBs.
Get your copy today!

When the ‘Pop-Up’ Store Sticks Around

Retailers have started to rethink their sprawling storefronts. Instead, companies are turning to smaller, more specialized locations that that can adapt to declining store revenues while addressing some new opportunities in selling to a connected consumer.

Twitter

© 2014 Street Fight.

Powered by WordPress. Hosting by Page.ly