NB: This is the fourth in a series of guest articles by Robert Cole, founder of hospitality and travel marketing strategy and technology consulting company RockCheetah. He also blogs at Views from a Corner Suite.
As publishers, consumers and law enforcement get more sophisticated, so will the bad guys. It is doubtful fake reviews can ever be fully eradicated.
The problem is that even amateur FROs realize that encouraging positive review posts about an establishment only goes so far…
The distance separating competitors can hypothetically be doubled if similar negative efforts can influence a comparable fall in a competitors ranking.
Those electing to go negative on competitors can be problematic. For their targets, FROs have the same impact as griefers found in online gaming communities. In gaming, griefers serve no purpose other than giving others grief. They may join a game, kill off members their own team and get banned as a result.
However, if a new profile is readily available, the griefer is free to join another game, or if hell-bent on retribution, join the same team using that new profile.
Due to the inclusive nature of hotel review and rating sites, FROs, like griefers, are free to behave as digital Rasputins, surviving repeated extermination attempts. And as any B-movie horror fan can confirm, extermination is a lot harder if there is an army of mutant zombies on the rampage.
Gaming platforms found the only way to effectively eliminate griefers was to start mandating registration, assessing participation fees, and/or allowing players to interact solely within a self-defined circle of friends.
While somewhat effective, the result is a reduced community size, with more limited breadth of engagement. The game may be a bit smaller, but it still beats playing FRO "whack-a-mole".
Casting a net(works)
Proactively inviting guests to post reviews and simplifying the process only goes so far. One can imagine as a hotel successfully builds positive sentiment through white-hat methods, a competitor using black-hat methods can simply turn up the heat by generating proportionally more negative reviews.
To fight against the expansion of fake reviews, the underlying technologies that drive the capture, collation and presentation of product reviews must be enhanced. The first question is if the reviewer is a real person; the second, did they actually experience the hotel.
The groups attempting to legally compel TripAdvisor to validate hotel stays for each reviewer are partially right – validating reviewer identities is a good idea, but they are completely wrong by suggesting that the review site should be solely responsible for confirming author identities or matching a reviewer to a specific hotel stay.
As the number of fake reviews grow and the sophistication of FROs increase, review sites will need to develop enhanced identity validation methods to maintain relevance. Tracking IP addresses and posting patterns is no longer enough. The review sites need to innovate to enable the development of and migration to review platform 2.0 technologies.
A similar challenge faced search engines a decade ago when they targeted the black-hat SEO pioneers. Those efforts were later enhanced to combat content farms that sprung up as work-arounds to the first waves of web spam filtering.
Reviewer identity validation involves both internal and external enhancements. Internally, review sites need to enable more comprehensive reviewer profiles.
Currently, TripAdvisor has a relatively weak profile with an old-school advertising/demographic focus on age, gender and metro (airport) location in addition to a name and email address. Some light travel preference information is also collected, but it does not assist in the validation process.
TripAdvisor implemented Facebook Connect in 2010 and later enhanced its integration Facebook’s Open Graph with Instant Personalization to offer maps based on its Cities I’ve Visited content and present lists of friends that have visited various destinations.
TripAdvisor smartly used the Facebook integration to also clean up its email database – it required the email used by TripAdvisor to match the email used by Facebook. This provided the dual benefit of helping clean up the TripAdvisor email database and confirming that the email was shared by a Facebook account.
While the Facebook connected users represent only a subset of TripAdvisor users, TripAdvisor reportedly experienced an opt-out rate below 1%. Yelp is similarly connected with Facebook as are Bing, Rotten Tomatoes and Pandora.
All of these companies are very interested in understanding consumer preferences and behavior – primarily to produce more relevant discoveries via search or more pertinent recommendations.
Advertising driven social networks want to provide marketers with insights into consumer purchase intent. By linking with more major social networks, TripAdvisor can help create a tighter net to help law enforcement track FROs.
Facebook and now Google, through Google+ are adamant to make certain human profiles are isolated from brand or fictional pages. It is those unique human profiles that TripAdvisor needs to tap into in order to create problems for the purveyors of fake reviews.
Apple and Amazon have extensive data repositories of profile and preference information, including credit card data for virtually all their users. Due to their transactional orientations, they would be less inclined to share customer validation data to assist third party review sites validate identities.
Twitter is littered with profiles known only by aliases and LinkedIn, with its business/employment orientation might not be an ideal best fit, but the more validation sources makes available, the more difficult it is for FROs to create and manage phony identities. Specifications such as OpenID 2.0 and OAuth 2.0 help facilitate the process.
Most importantly, each of these externally validated links provides an additional tether that may potentially be linked back to the ultimate author of a review. Triangulating across multiple third party validation sources also helps make concealing the ultimate review source more difficult.
Linking relationships between sockpuppet profiles across social networks to artificially boost authority becomes highly risky if a single profile is revealed to be fictional through the course of an investigation.
The above provides the first line of defense against FROs – similar to putting an alarm system on a house. While accomplished criminals will still be able to strike, it should serve as a deterrent for those less skilled and or motivated.
Traxo has the right idea
Validating stay data makes a FRO’s life truly miserable. The challenge is how to put this in place without infringing reviewer privacy or compromising the integrity of hotel guest history records.
The best solution is for the review sites to develop an opt-in verified reviewer program that allows users to provide third party frequent guest and/or booking site details that can validate hotel stays.
Traxo has already laid the groundwork for such a program and provides perhaps the most elegant process as no action is required on the part of the hotel company or hotel guest once a guest is registered.
Here is how it works: users securely provide frequent guest usernames and passwords to Traxo. Then, Traxo periodically interrogates the accounts and synchronizes stay information with its servers.
Traxo automatically sends "Welcome Back" emails following the trip that allows the traveler to grade the travel supplier (it works for airlines, and cars as well) on a 1-5 scale. For a review site, all that is necessary is for a verified reviewer program participant to enters the review.
There is also the opportunity to validate stays by booking source. For a trip taken last month, Traxo saw that I had stayed at a Marriott through my Marriott Rewards account on Marriott as well as Travelocity (the booking source)
Utilizing this approach, the review site benefits significantly by dynamically querying its group of Verified Reviewers immediately following trips. The hoteliers also benefit from having the review site post legitimate, verified reviews.
From the reviewer’s perspective, the review process is greatly simplified. Plus, an opportunity exists for reviewers to maintain their anonymous review site profiles – identities confirmed as part of a back-end process would not need to be exposed to the general public.
Interestingly, Traxo already takes this approach one step further and provides a so-called "Travel Score" - similar to a Klout "Authority" score – that is calculated by an algorithm that considers total miles and days traveled, loyalty program status and the geographic distribution of trips.
Additionally, Traxo weights verified and recent trips higher than manually entered or older trips. One can see these factors logically apply to travel review sites – particularly the extra weighting for validated reviews.
One additional validation enhancement can be seen on Rotten Tomatoes. The movie review site has segregated its reviews between critics and the general audience. Site visitors are able to share and participate in the community, but are also able to contrast the differential between the pros and the amateurs.
Rotten Tomatoes categorizes reviewers further by flagging top critics separately from other critics. Frequent contributors among audience members are identified as "Super Reviewers", again providing greater credibility for validated profiles.
The site is able to work with the top critics because only a small indicative excerpt from the review is revealed, with a link to the author’s website for the full review. As a result, the site serves as a hub that helps create exposure for critics, as well as a source of site traffic for the reviewer.
A similar approach by a travel review site could potentially leverage travel guides, travel writers and bloggers to accomplish the same objective – while simultaneously reducing the impact of FRO’s intent on gaming the ratings to benefit their clients.
Getting personal
Another way to stymie the FROs is to increase the personalization of the review site.
One of the primary criticisms of TripAdvisor is that it provides a homogenized perspective of the hotels – it provides many filter criteria to trim down the hotel list based on various criteria, but the filters are property-centric and the new result set includes the same list of reviews. Reviews may only be filtered by family/couples /business to reduce the number of reviews to scan.
While the Facebook integration definitely helps provide perspectives of trusted friends, there are opportunities to apply more semantic and collaborative analytical processes to produce more relevant reviews.
Every traveler essentially seeks the same result – the hotel that represents the best experience/price ratio, based on the specific needs for a particular itinerary. But it needs to go further – it’s all about context.
Does a hotel match an individual’s sense of style? Does it match up favorably with other hotels the guest liked when traveling on a similar itinerary? Providing answers to these questions not only provides a much more relevant hotel recommendation, but also introduces sufficient nuance to frustrate even the most accomplished FROs.
The solution is to enable a more semantic and persona-based context into the hotel reviews that may be considered when ranking candidate hotels. Applying semantic search processes should greatly assist travelers by organizing the daunting amount of text provided by an ever growing army of reviewers.
However, fake reviews can wreak havoc on the sentiment-related context. Fortunately, this problem can be remediated by introducing a second contextual dimension – relative context.
If hotel ratings and reviews can be mapped through the lens of people with similar tastes, the resulting rankings should become even more relevant to the guest. Amazon uses its collaborative filtering technology to successfully create recommendations based on people with similar interests and purchase histories.
By combining semantic analysis with collaborative filtering, FROs will be frustrated by the greater granularity of factors that drive common interests among hotel guests. Additionally, the combination also negates the impact of artificially polarized ratings. Creating opposing reviews of competitors only serves to disassociate the client property from its natural competitive set.
Logically, if it is made to look like the type of people who love Hotel A hate Hotel B, Hotel A will be less likely to attract business from Hotel B’s prospective guests. Most FROs are fundamentally interested in attracting guests that might be more inclined to patronize a competitive hotel.
Finally, integrating the semantic and collaborative filtering with an individual’s social graph will further help to structurally filter out bogus reviews from consideration as reviews more closely associated with a social graph should be granted stronger signal strength than reviews originating from sockpuppet accounts that are not associated with human social graphs.
The solutions proposed will not eradicate phony reviews, but they will be a deterrent. Ideally, FROs will choose to avoid hotel review sites and direct their attentions toward easier prey.
Social networks including Facebook, Google+, Twitter, LinkedIn and travel specific networks will become increasingly important – not necessarily as the review billboards they now resemble, but as a method to validate decisions based on the experience of legitimate Friends of Friends serving as subject matter experts.
Engaged groups of friends and acquaintances will be using the social web to interact with brands, pundits and those flagged as knowledgeable in travel by on that traveler’s social graph.
As the sophistication of review sites increases, those employed in the Fake Review Optimization business will be challenged to demonstrate knowledge or expertise and have a very difficult time penetrating human social graphs in a meaningful way.
As the battle of good versus evil rages in the travel review space, the most important question a hotel guest can ask is "Who can I trust?"
In the not so near future, that may turn out to be someone they have not personally met, but somebody trusted in turn by a friend. Or a social networking savvy travel agent.
Perhaps the best advice for travelers relying on user generated reviews and ratings comes from Harry Potter author J K Rowling when she writes:
"Never trust anything that can think for itself if you can't see where it keeps its brain."
The concept of Caveat Emptor, translated from Latin, "let the buyer beware" first arose in the early 16th century. Despite the amazing technological advancements over the past 500 years, fundamental human nature has not changed. If the next 500 years will be any different is clearly a topic best left open to debate.
If there is money to be made, there will always be parties interested in profiteering from misleading consumers. As a result, the extinction of FROs is unlikely.
Regardless, it seems like a good time to start putting those wearing black hats in the review space on the endangered species list.
NB: This is the fourth in a series of guest articles by Robert Cole, founder of hospitality and travel marketing strategy and technology consulting company RockCheetah. He also blogs at Views from a Corner Suite.
NB2: These are the earlier articles...