Ask HN: How to create certified or trusted reviews?
3 comments
are you looking for a product that you can integrate with your site now? or you want an algorithm?
Filtering "fake or deceitful reviews" is a complicated task to accomplish. While you won't find the exact solution, you can research how companies like Yelp and AngiesList have done it.
There are many factors in the formula, but a lot of it revolves around looking at who the user is exactly. By looking at how often they contribute, how long they have been a member of the site, and incorporating a rating system into the review (helpful/not helpful) you can begin to determine which are the best reviewers.
Verifying the user is real is another step you can take. You can do this by adding a verification step into the account creation process (i.e. responding to a email) prior to submitting a review. Setting default time limits to how long the account has to be active before they can make a review also can work (this avoids the random user in the heat of the moment writing a false review or one that is not fully thought out). If your users have "profiles" in the system, similar to Yelp, you can give more credibility to a user who takes the time to fill out their profile info.
Another method is to track IP addresses and compare to where the place of business being reviewed is. Although this will not work in all cases (i.e. someone traveling or logging in from home for a vacation experience), it does lend some credibility.
By giving each method you incorporate into your system a weight, you can then add the weights up to come up a validity indicator. Each method could be weighted on a scale of 1 to 10 and the higher the weight awarded for that method the more credible. You then have to weight the methods in general. In the end the formula combines all those weights together to get a total weight which should give you an idea of how credible the review is.
There are many factors in the formula, but a lot of it revolves around looking at who the user is exactly. By looking at how often they contribute, how long they have been a member of the site, and incorporating a rating system into the review (helpful/not helpful) you can begin to determine which are the best reviewers.
Verifying the user is real is another step you can take. You can do this by adding a verification step into the account creation process (i.e. responding to a email) prior to submitting a review. Setting default time limits to how long the account has to be active before they can make a review also can work (this avoids the random user in the heat of the moment writing a false review or one that is not fully thought out). If your users have "profiles" in the system, similar to Yelp, you can give more credibility to a user who takes the time to fill out their profile info.
Another method is to track IP addresses and compare to where the place of business being reviewed is. Although this will not work in all cases (i.e. someone traveling or logging in from home for a vacation experience), it does lend some credibility.
By giving each method you incorporate into your system a weight, you can then add the weights up to come up a validity indicator. Each method could be weighted on a scale of 1 to 10 and the higher the weight awarded for that method the more credible. You then have to weight the methods in general. In the end the formula combines all those weights together to get a total weight which should give you an idea of how credible the review is.
Okay - here's where my mind went when presented with this problem: be like an ombudsman. When I call up the TIO (http://www.tio.com.au/) to complain about my mobile phone supplier, the first thing they do is give me a reference number and a phone number at the company that is reserved for "high level" customer service enquiries (ie. I get to speak to a "case manager" rather than having to navigate the layers of support from the ground-up).
So give the company that the review is about a chance to respond prior to publishing the bad review. If they don't respond, publish it. If they do respond, then you can still publish the bad review but the OP may edit it so that it's not quite so caustic.
This could be regulated quite easily by requiring the OP to provide a valid contact email address and URL for the company about which they're providing a review.
Or you could do like the TIO does and provide the customer with a reference number to contact the company they originally dealt with and allow the company to login and use that same reference number to tell their side of the story.
Your company is then like, OmniOmbudsman :) (.com ... take it, it's available)
So give the company that the review is about a chance to respond prior to publishing the bad review. If they don't respond, publish it. If they do respond, then you can still publish the bad review but the OP may edit it so that it's not quite so caustic.
This could be regulated quite easily by requiring the OP to provide a valid contact email address and URL for the company about which they're providing a review.
Or you could do like the TIO does and provide the customer with a reference number to contact the company they originally dealt with and allow the company to login and use that same reference number to tell their side of the story.
Your company is then like, OmniOmbudsman :) (.com ... take it, it's available)
The issue that we are struggling with is the prevalence of fake or deceitful reviews across the web. Does anyone have any suggestions/ideas/products that could help us in creating a trusted system?
Thanks