Automotive Digital Marketing ProCom

Professional Community for Car Dealers, Automotive Marketers and Sales Managers

Loading... Be Patient!

How Does Google Identify Fake Online Reviews?

In this blog post I reference two published research papers that cover a very important topic: Identifying fake online reviews. The two research papers I reference below, start to shed light on how online communities and the search engines will combat review spam.  The second research paper listed in this article includes a researcher from Google; Natalie Glance.  

 

As Google increases its consumer influence through Google Places, making sure that the reviews posted on Google Places are authentic is critical.  Review sites like Yelp and TripAdvisor have much at stake as well especially with national advertisement running from companies like Reputation Defender or Reputation.com. 

 

Research Paper #1

 

The first published research paper I will reference was completed by Myle Ott, Claire Cardie, and Jeff Hancock from Cornell University.  Their work gives us insight into the growing interest in assuring that online reviews are authentic  This topic has been discussed many times within the automotive community, and I thought this research would appeal to many readers.

 

Opening  Abstract

 

Consumers' purchase decisions are increasingly influenced by user-generated online reviews [3]. Accordingly, there has been growing concern about the potential for posting deceptive opinion spam- fictitious reviews that have been deliberately written to sound authentic, to deceive the reader [15].

But while this practice has received considerable public attention and concern, relatively little is known about the actual prevalence, or rate, of deception in online review communities, and less still about the factors that influence it.

We propose a generative model of deception which, in conjunction with a deception classier [15], we use to explore the prevalence of deception in six popular online review communities: Expedia, Hotels.com, Orbitz, Priceline, TripAdvisor, and Yelp.

We additionally propose a theoretical model of online reviews based on economic signaling theory [18], in which consumer reviews diminish the inherent information asymmetry between consumers and producers, by acting as a signal to a product's true, unknown quality.

We did find that deceptive opinion spam is a growing problem overall, but with different growth rates across communities. These rates, we argue, are driven by the different signaling costs associated with deception for each review community, e.g., posting requirements. When measures are taken to increase signaling cost, e.g., filtering reviews written by first-time reviewers, deception prevalence is effectively reduced.

 

Download PDF of Research

 

To download a free copy of the research paper, click on the link below to get the PDF document:

 

Estimating the Prevalence of Deception in Online Review Communities

 

Research Paper #2

 

A second research paper entitled "Spotting Fake Reviewer Groups in Consumer Reviews" is also a great read and the research was conducted by Natalie Glance from Google, and Bing Liu and Arjun Mukherjee from the University of Illinois at Chicago.

Opening Abstract

 

Opinionated social media such as product reviews are now widely used by individuals and organizations for their decision making. However, due to the reason of profit or fame, people try to game the system by opinion spamming (e.g., writing fake reviews) to promote or demote some target products. For reviews to reflect genuine user experiences and opinions, such spam reviews should be detected. Prior works on opinion spam focused on detecting fake reviews and individual fake reviewers.

 

However, a fake reviewer group (a group of reviewers who work collaboratively to write fake reviews) is even more damaging as they can take total control of the sentiment on the target product due to its size. This paper studies spam detection in the collaborative setting, i.e., to discover fake reviewer groups. The proposed method first uses a frequent itemset mining method to find a set of candidate groups.

 

It then uses several behavioral models derived from the collusion phenomenon among fake reviewers and relation models based on the relationships among groups, individual reviewers, and products they reviewed to detect fake reviewer groups.

 

Additionally, we also built a labeled dataset of fake reviewer groups. Although labeling individual fake reviews and reviewers is very hard, to our surprise labeling fake reviewer groups is much easier. We also note that the proposed technique departs from the traditional supervised learning approach for spam detection 

 

Download PDF of Research

 

To download a free copy of the research, click on the link below to get the PDF document:

 

Spotting Fake Reviewer Groups in Consumer Reviews

 

Reputation Management and Marketing

I hope you found this research helpful.  I have to dig into the meat of the research more before formulating my feedback.

 

If your business has been struggling with an online review strategy, I invite to you to join the discussion at the upcoming 2012 Automotive Boot Camp.  A number of workshops will be offered to discuss this important topic. 

 

 

Brian 

 

Brian Pasch, CEO

 

brian@pcgmailer.com

 

Views: 863

Tags: fake reviews, google places reviews, google reviews, online reviews, reputation management

Comment

Oops... You need to stop "Lurking" on ADM and become a more genuine Automotive Professional by completing your membership registration. As a registered ADM Member, you can post comments, publish your own articles (be a star!) and start Forum discussions. Stop being an online "Peeping Tom" and JOIN ADM RIGHT NOW!

Join Automotive Digital Marketing ProCom

Comment by JVRudnick on May 7, 2012 at 2:03pm

Been on this topic now for years....here's a link to our original blog post on same - fake reviews for car dealerships - http://www.canuckseo.com/index.php/2010/08/online-maps-review-credi...

 

And from what I see, while the methodology has changed somewhat, there are still TONS of fake reviews being used...sigh.....just hope Google catches them all, eh!

Comment by Jennifer Paine on April 30, 2012 at 9:43am

Great post!!!!  Thank you for the info.  Our dealership works so hard at getting legitimate and honest reviews from our customers.  I just get so frustrated when I see others who are obviously scamming the system.  Here's to hoping they all get caugt.  I agree with Kevin,  (yes, I do know who Tony Baretta is...lol)...If you can't do the time....don't do the crime.

Comment by kevinkelly.it on April 30, 2012 at 7:37am

Great post !  Perhaps the only importance we really need to understand is that the many dealerships that post fake online reviews, positive for them, negative for the competition, can and will be caught.  The consequences is where we should place the emphases.    SERP results limited to page 3 is bad, Google death sentence is a dealership death sentence.  In the famous words of Tony Baretta, don't do the crime, if you can't do the time.  ( most likely 80% reading this are too young to know Tony Baretta) 

Comment by Marc Bodner on April 30, 2012 at 6:22am

Great post Brian.  Been a while since I had to do that kind of math; wasn't much better at it now! However since I make my living working with dealers on building better reputations it's good to see the topic getting this much play.  

The sad fact is too many dealers farm out management of this area to third parties.  They see it as important but don't want to take internal time to manage this process.  Reputation management, if executed properly to achieve goals tied to leads and sales, needs to be managed at the dealership with the customer.  Data mining past consumers and soliciting reviews usually comes with an incentive; return this survey for a free oil change.  That practice is now under scrutiny by EVERY third party site.  

Work at the dealership with your customers and staff, and the online reputation will grow in a professional and honest manner. 

Comment by John Weltz on April 30, 2012 at 5:39am

Thanks Brian. This info is crucial. You are ahead of the curve again!

Comment by Tom Gorham on April 30, 2012 at 5:22am

Brian, thanks for this post on one of my favorite subjects.  I do believe that Google is preparing to crack down increasingly on fake reviews and that some dealers will be penalized for things they have been doing all along.  After alll, it's Google's reputation too.  They have strong incentive to make sure their reviews are trustworthy.

Comment by Kevin Frye on April 30, 2012 at 5:02am

Great article Brian, and one of my pet-peeves. In fact, buying fake dealership reviews was one of the things I shared at my DD12 Presentation "The 7 Highly Ineffective Habits of Internet Buffoons". I shared several screenshots of dealerships who are blatantly buying fake reviews and how to detect them. As per Google Places, they have an advantage to finding fake reviews - they can look at the IP addresses of where the reviews are coming from. When your dealership is in the United States, but 99% of your reviews (all 5-star by the way...) have an IP address in India, Google can see that. In most cases, if you report the reviews to Google, they will complete the audit and catch the culprit.

With all that said, my frustration comes from working incredibly hard for several years to build and maintain an outstanding online reputation. That means both great and negative reviews (that are opportunities for your dealership to learn and become better). Unfortunately we have many other dealers who figure the easy way to build an online reputation is to buy one, and these are the same folks who continue to give our industry a bad name. 

Sadly, this is a problem in many other industries as well. I agree with these efforts in identifying fake reviews and look forward to these policies exposing those who are buying or writing fake reviews...

Comment by Mark Dubis on April 29, 2012 at 6:23pm

As long as there are surveys you will find people trying to manipulate surveys.  A classic example are the current factory survey programs which I predict will become obsolete within two years due to the voice of the consumer sharing their opinions online and in various social channels.   Since there have been numerous discussions previously about the many problems with factory surveys I will not address them here.   The consumer response rate for these surveys continues to spiral downward.

Manipulating, hiding and removing bad reviews for dealers have become big business for many companies offering “Social Media” services.  As Joe Bessimer said, “There is a sucker born every minute.”  If dealers are willing to pay thousands a month to remove bad reviews, these companies will only be too happy to take their money.

Consumers are discovering and many have already posted comments on sites like “PissedConsumer.com” that some dealer review sites are not consumer accurate.  Up till now most solutions allowed dealers to hide or “kill” negative reviews and unfortunately dealers are still drawn to this model.  The problem is when a consumer sees testimonials either on a dealer’s website or on a third party site and there are NO negative or neutral reviews, the consumer discounts the validity of all the reviews there. 

Think about it.  You visit a car dealer’s website and click on the testimonial page where you see reviews the dealership posted and they are all glowing.  So we are saying to the customer, “Trust me, I’m a car dealer and believe all the info on MY website because it’s true.”   Yeh, right!!

Any dealer review platform that is offered by a website vendor or a classified site is suspect as their revenue comes from that dealer specifically to promote their business and vehicle sales. Cobalt now offers review microsites that have a “Confirmed Customer” highlight to provide validity, but again it’s on the dealer’s own site. When I see 253 confirmed reviews and not one of them is under 4 stars, it says “bogus” to me and will say the same to consumers. 

After 20+ years in the business and after working with dealers and Internet Managers while I was the editor at Digital Dealer I, like you Ralph and Brian have found most people in this industry are good, hardworking people who do the right thing every day.

I got upset when I saw good dealers getting slammed on these review sites and wanted to do something to help dealers and sales professionals while at the same time providing accurate information (un-manipulated) for consumers.   When we created Carfolks we spoke to over 100 GM’s and Dealer owners who shared their concerns, frustrations, and their goals when it came to online reviews.  They had no issue with negative reviews as long as they could respond and work with the customer to fix the issue, and they wanted a way to make their people responsible for the customer relationship. 

Our entire site and business model was designed from the ground up, with no influence from “our website profits” (we have no website, DMS, or CRM company) so our only mission was to help portray dealers in an accurate fashion and make it simple for dealers to engage their loyal customers and leverage their comments in a focused marketing program. 

Carfolks has been slow to ramp up but dealers are finding out quickly how powerful and easy our program can be and dealers on our platform will be starting to quickly standout against their competitors in local markets.   We also didn’t like the “pay to play model” so we give a free page to every car sales person in the country to build their personal brand, and they can take their reviews with them to any dealership in the US or Canada.  That means they have the tools to really be successful.

Good dealers have nothing to hide and should be proud of their people, as they do a great job day in and day out.  Now they need to be given the tools to take that professionalism to the next level.

Mark Dubis
co-founder Carfolks.com

Comment by Ralph Paglia on April 29, 2012 at 3:23pm

Brian - I just finished downloading and reading both of the PDF files you provided links to... They BOTH made my brain ache! The mathematical application of formulaic approaches to identifying fake reviews is fascinating, and I believe the first document you reference has a lot of value, albeit too academic for most members of this community.  However, I have a fundamental disagreement with the assumption sets used in the second research report co-authored by Google associates.  Applying the assumption that if a single review is too far different from all the other reviews in aggregate, thus identifying fake reviews is fundamentally flawed because it is based on the assumption that the services delivered by merchants are uniform and consistent.  As we all know, this is simply not true... In the case of car dealerships, a consumer's experience in the F&I department may be considerably different depending on which F&I manager processes their contract and paperwork.  Any time a mathematical formula is applied to review evaluations that is based on the assumption that a given dealership or merchant delivers consistent customer experiences, it is doomed to being incorrect in the real world.

Let me further explain my thinking by using the opposite... If we are talking about movie reviews, then the algebraic and deviation based approaches to identifying fake reviews MAY hold some degree of validation because it is the SAME MOVIE that everyone sees, regardless of the location they chose for viewing the movie... Alternatley, even when reviews of a given dealership are all based on the same location, the experience that is delivered to customers varies widely depending on a number of factors, especially because they may have dealt with a different group of 5 people in their vehicle purchase process than other consumers buying vehicles at the same dealership.

So, despite the great scientific rigor that both of these document you reference seem to apply, their assumptions on consistency of the consumer experience may be less valid for businesses such as car dealerships where the consumer experience truly does vary from one day to the next and from one consumer to the next, depending on a mix of variables.

Comment by Keith Shetterly on April 29, 2012 at 2:21pm

@ Ralph:  Right on the money.

@ Brian:  Excellent post!!!!!!!!

ADMPC is a Network for Car Dealers, Automotive Marketing, Advertising and Management Pros sharing Digital Strategies and Execution Tactics.

Please Consider Automotive Marketing Professional Community Sponsors

ADM Badge

Loading…

Based On Your Interests...

Automotive Marketing Tools

Get ADM Toolbar

Click here to take the ADM Member Survey

 

Google Automotive Network Targeted Placement Ads

Getting too many emails from ADM? Click mailbox below to control which types of alerts and updates you are sent......


Instagram

© 2014   Created by Ralph Paglia.

ADM Badges  |  Report an Issue  |  Terms of Service