There are many fake reviews in TripAdvisor. It is easy to spot one that harvests fake reviews. They are usually reviewers just have reviewed only the current one with a 5*. This is not to say that all the once-only reviewers are fake. But if a place is reviewed all 5* by these newbies, that should ring a bell. Moreover, I have seen in freelancing sites "employers" asking for 100s of reviews in sites such as TripAdvisor for a hundred dollars! It can also be friends filling TripAdvisor reviews for a restaurant owner, or even the restaurant employees creating multiple fake accounts to accumulate 5* ratings.
It can also be social engineering. When I stayed in New Delhi in 2012, the hotel owner asked me to rate them 5* in TripAdvisor. That was my first review! I guess a normal person would have given 5*. Does not hurt. right? But I gave 4*, the rating I thought the hotel deserves (than what the owner hinted)! Would be easier to manipulate one's rating with some benefit (like a discounted rate) though. I am not going to give fake 5* ratings for free stuff though. :) I hope.
Always check the 1* ratings. They may show how the place mistreats the visitors. Also 4* ratings. They tend to be the genuine happy customers. Fake reviewers always give 5*. Not 4*. No one pays to give 4* ratings! Similarly, 2* ratings tend to be more rational than 1* too, unless the restaurant is really fishy and unpleasant.
I sent a detailed report to TripAdvisor content-integrity sometime back on a sample situation. Their response was along the lines of "At the same time, we have privacy policies in place which prevent us
from being able to divulge to anyone external to TripAdvisor the results
of any investigations."
Apart from being Nepal restaurants in Europe, what is the other similarity shared by "Sushi King, Wijnegem", "Sushi King, Malle", "Leo, Lisbon, Portugal", and "Fishtail, Lisbon, Portugal"?
They all have been reviewed by the same set of early reviewers who gave
5* reviews with extra-ordinary positive reviews. These reviewers also
created their accounts just to review these restaurants!
The first 2 restaurants are in Belgium and as their name suggests, are owned by a same chain. The last 2 are in Portugal, and owned by an individual (see the attachment for proof, taken from TripAdvisor). There are many accounts that were created to review 2 or 3 among these or just 1 of these restaurants! Some of these reviewers have reviewed 1 or 2 of these restaurants more than once! (5-star each time).
I have attached two reviewers with this association for these 4 restaurants. There are many. Just go through each of these restaurants and click the 5-* ratings. You can see reviewers created solely to review one or a few of these 4 restaurants. All these reviewers are Nepalese or Indian.
The first 2 restaurants are in Belgium and as their name suggests, are owned by a same chain. The last 2 are in Portugal, and owned by an individual (see the attachment for proof, taken from TripAdvisor). There are many accounts that were created to review 2 or 3 among these or just 1 of these restaurants! Some of these reviewers have reviewed 1 or 2 of these restaurants more than once! (5-star each time).
I have attached two reviewers with this association for these 4 restaurants. There are many. Just go through each of these restaurants and click the 5-* ratings. You can see reviewers created solely to review one or a few of these 4 restaurants. All these reviewers are Nepalese or Indian.
One can do an association rule
mining across all these profiles and get more associations to show how
these reviewers are having a strong correlation with each of these.
These 4 may be more than 4. I mean, if you find a 5th association, try
to build a cluster to see whether they share the same pattern. I am just
a volunteer TripAdvisor member. I gave up after finding 4.
1. It is highly unlikely that a large share of reviewers visited all 4 of these restaurants in 2 different countries, and created their accounts to review only these 4!
2. I suspect they have some incentives.
2.1. I have previously seen random employers in freelancer websites seeking freelancers to bulk review their pages in Facebook, TripAdvisor, etc..
2.2. They are probably friends of the owner. In case of Leo, it is currently #1 in Lisboa with just 136 reviews. These reviews can be from bots. Even if they are from real humans, it is not hard to find 136 "friends" for an Indian/Nepalese restaurant owner in Portugal.
3. Given that Leo is the #1 in Lisboa now, unsuspecting tourists visit there since it is a "must" now, as listed #1 by TripAdvisor. Some of them are satisfied naturally. The food is not bad - and deserve 3* - 4* anyway. So they leave a positive review. Some were a bit disappointed learning this is just yet another Indian restaurant. However, the initial bootstrap from the fake reviewers were strong enough for Leo.
As a continuous volunteer reviewer in TripAdvisor, I am disappointed to learn that TripAdvisor's approval workflow is not smart yet.
1) The ranking algorithm should be changed to
1.1) ensure a meritocracy than a democracy. In the Internet, each account is not necessarily a human. So democracy in such systems is flawed. Currently, there is no difference getting 100 5* reviews from fake accounts or bots and 100, real reviewers with proven track record.
1.2) give more importance to the number of reviews. 1000 4* reviews are probably better than 4.5* rating with 137 reviews. This should of course not harm the new businesses. So needs a smart approach. Requires further research. Currently this is the loophole how Leo managed to become the first in Lisboa!
2) The approval workflow should consist of a data mining approach to ensure reviews are not creating a pattern. As of now, I was able to find a pattern very easily among these companies. With a large set of engineers and the management interface/API, this should be simple for TripAdvisor team.
3) Fix the reporting system. Currently it is a bot. First, I report, and a canned response asked me to send an email to this address. I am not even sure whether this will be considered properly.
Leo coming to the top spot reminds me the story - http://www.independent.co.uk/l
The difference is Leo actually exists as a normal Indian/Nepalese restaurant. But there are much better Nepal or Indian restaurants around, and this surely is not the #1 of Lisboa.
These first 100 reviews were given by fake reviewers to give an initial bootstrap to secure the first place, and following were social engineering. The tourists who blindly follow TripAdvisor and pay a visit here assuming this to be the best of Lisboa. Following this huge popularity, this small canteen restaurant is even unable to cope with the number of customers, making the waiting time grow large. Unsuspecting customers take all these positively. Some even fell for the mediocre Indian food, which can also be found in any other Indian/Nepal/Bengal restaurant around.
1. It is highly unlikely that a large share of reviewers visited all 4 of these restaurants in 2 different countries, and created their accounts to review only these 4!
2. I suspect they have some incentives.
2.1. I have previously seen random employers in freelancer websites seeking freelancers to bulk review their pages in Facebook, TripAdvisor, etc..
2.2. They are probably friends of the owner. In case of Leo, it is currently #1 in Lisboa with just 136 reviews. These reviews can be from bots. Even if they are from real humans, it is not hard to find 136 "friends" for an Indian/Nepalese restaurant owner in Portugal.
3. Given that Leo is the #1 in Lisboa now, unsuspecting tourists visit there since it is a "must" now, as listed #1 by TripAdvisor. Some of them are satisfied naturally. The food is not bad - and deserve 3* - 4* anyway. So they leave a positive review. Some were a bit disappointed learning this is just yet another Indian restaurant. However, the initial bootstrap from the fake reviewers were strong enough for Leo.
As a continuous volunteer reviewer in TripAdvisor, I am disappointed to learn that TripAdvisor's approval workflow is not smart yet.
1) The ranking algorithm should be changed to
1.1) ensure a meritocracy than a democracy. In the Internet, each account is not necessarily a human. So democracy in such systems is flawed. Currently, there is no difference getting 100 5* reviews from fake accounts or bots and 100, real reviewers with proven track record.
1.2) give more importance to the number of reviews. 1000 4* reviews are probably better than 4.5* rating with 137 reviews. This should of course not harm the new businesses. So needs a smart approach. Requires further research. Currently this is the loophole how Leo managed to become the first in Lisboa!
2) The approval workflow should consist of a data mining approach to ensure reviews are not creating a pattern. As of now, I was able to find a pattern very easily among these companies. With a large set of engineers and the management interface/API, this should be simple for TripAdvisor team.
3) Fix the reporting system. Currently it is a bot. First, I report, and a canned response asked me to send an email to this address. I am not even sure whether this will be considered properly.
Leo coming to the top spot reminds me the story - http://www.independent.co.uk/l
The difference is Leo actually exists as a normal Indian/Nepalese restaurant. But there are much better Nepal or Indian restaurants around, and this surely is not the #1 of Lisboa.
These first 100 reviews were given by fake reviewers to give an initial bootstrap to secure the first place, and following were social engineering. The tourists who blindly follow TripAdvisor and pay a visit here assuming this to be the best of Lisboa. Following this huge popularity, this small canteen restaurant is even unable to cope with the number of customers, making the waiting time grow large. Unsuspecting customers take all these positively. Some even fell for the mediocre Indian food, which can also be found in any other Indian/Nepal/Bengal restaurant around.
No comments:
Post a Comment
You are welcome to provide your opinions in the comments. Spam comments and comments with random links will be deleted.