Live data from Hacker News

Some Amazon Reviews Are Too Good to Be Believed – They're Paid For

npr.org

41–50 of 273 posts

Re: Some Amazon Reviews Are Too Good to Be Believed – They're Paid For

#41
post #33
post #20

Earlier quoted context omitted.

You'd think that Amazon would have the machine learning resources to detect likely-fraudulent items with high accuracy, should they be interested in preventing such obvious fraud.

> You'd think that Amazon would have the machine learning resources to detect likely-fraudulent items with high accuracy, should they be interested in preventing such obvious fraud. They must not be interested. After all, they still get paid and have a level of legal insulation due to how they've structure Amazon marketplace. I wonder how long it'll take for their overall brand to take a hit over this, though.

There's probably a huge number of shoppers that use Amazon for nearly everything, myself included, that would just toss the cheap crappy thing they got off Amazon and try another brand from, you guessed it, Amazon.

I don't know how many horrible sets of bath rugs and towels I went through on Amazon before I finally bit the bullet and bought an actual expensive brand and the difference is staggering. Before someone asks they're from Frontgate (not on Amazon).

Point being I'm just sending Amazon money over and over again.. It's definitely made me go back to stores where I can see the product for certain things.

Re: Some Amazon Reviews Are Too Good to Be Believed – They're Paid For

#43
post #23

There was a great Planet Money podcast on the genesis of ReviewMeta which I've been using on Amazon reviews since. https://www.npr.org/templates/transcript/transcript.php?stor...

Since listening to that podcast a few weeks ago I've also started checking both ReviewMeta and FakeSpot for every Amazon listing I think about buying.

It's really made shopping on Amazon a frustrating experience -- well over half the products I've checked (in a couple specific product categories) have had F or D ratings according to both sites. I've wondered if some of these are false positives (i.e. ReviewMeta flags a product's reviews as fake but they're actually legitimate), but when you look at their reasoning in the breakdown they give, it's usually pretty convincing that many/most of the reviews are indeed fake.

I feel like this might be a case where ignorance is (mostly) bliss ... when I naively thought the vast majority of reviews on Amazon were genuine, shopping was much faster and simpler, and I mostly had good experiences buying well-reviewed products despite that many of those reviews were probably fake.

Re: Some Amazon Reviews Are Too Good to Be Believed – They're Paid For

#44

It's common knowledge among most techies that the majority of Amazon reviews are fake and that counterfeit items are a big issue, but most of the general population is completely unaware. I was searching for an inexpensive pair of polarized sunglasses last week and the "Amazon Recommended" ones had an F on FakeSpot with 70%+ fake reviews.

How legitimate is FakeSpot though? I find nearly everything has an F with a high amount of fake reviews when I use it.

It's been reliable for me.

> I find nearly everything has an F with a high amount of fake reviews when I use it

Seems like a reflection of how bad the reviews on Amazon are, not a reflection of the quality of fakespot.

Re: Some Amazon Reviews Are Too Good to Be Believed – They're Paid For

#45

Star ratings in general are just poor. In a 1 to 5 star rating, 3 is average but most people would perceive 3 stars to be above average. People tend to leave reviews only when they are displeased with the product, really pleased with the product or get something in return for making the review, in this case free products. I tend to just filter on reviews that are 2 or 3 stars and use those to work out the cons of the…

From the distribution of reviews scores, even when separated by category, the majority of ratings given are 4-5 stars (from one of my 2017 analyses: https://minimaxir.com/img/amazon-spark/category_breakdown.pn...)

My full analysis of Amazon rating values is here (https://minimaxir.com/2017/01/amazon-spark/), although it may be obsolete and it's not possible to get better/more recent data anymore.

Re: Some Amazon Reviews Are Too Good to Be Believed – They're Paid For

#46

Star ratings in general are just poor. In a 1 to 5 star rating, 3 is average but most people would perceive 3 stars to be above average. People tend to leave reviews only when they are displeased with the product, really pleased with the product or get something in return for making the review, in this case free products. I tend to just filter on reviews that are 2 or 3 stars and use those to work out the cons of the…

Really? I think most people perceive anything less than 4 stars (or even 4.5 stars for some) as "bad". I for one won't buy anything that are "merely" 3-stars.

True, I gave a tech book 3 stars once and the author contacted me to request that I reconsider, explaining that anything less than 5 stars could doom a brand new tech book on Amazon. He offered personal support and training in return for raising the score. It didn't work, but it also convinced me that he has a good point regarding the score: Reviews are difficult ground due to their subjective nature, and a lot of reviewers are simply qualitatively bad at writing reviews. If an author just sits back and does nothing, they are risking their brand on the individuals who take the time to write reviews, and that set of individuals could be--even randomly--pretty bad.

For example: I've seen many, many 1-star "product was damaged in shipping" and similar e.g. "received wrong product" one-reviews for products that were absolutely stellar. Getting into even more tricky ground, I've seen 1-star "product is unscientific" reviews left on books containing information that is mainly intractable to science, like studies in history.

Re: Some Amazon Reviews Are Too Good to Be Believed – They're Paid For

#47

Earlier quoted context omitted.

: ) Well don't take it on trust, validate it works. And keep validating it still works, because one day it won't. Everyone seems to sell out in the end.

What do you mean by "validate it"?

I've used fakespot for purchases for a while now. I feel that it's helped me avoid bad products. Thus... it's validated (at lease for myself)

Re: Some Amazon Reviews Are Too Good to Be Believed – They're Paid For

#48
post #20

It's common knowledge among most techies that the majority of Amazon reviews are fake and that counterfeit items are a big issue, but most of the general population is completely unaware. I was searching for an inexpensive pair of polarized sunglasses last week and the "Amazon Recommended" ones had an F on FakeSpot with 70%+ fake reviews.

You'd think that Amazon would have the machine learning resources to detect likely-fraudulent items with high accuracy, should they be interested in preventing such obvious fraud.

Similarly for Kindle Unlimited despite the fact that Amazon clearly has the technology to address Markov Chain generated fakes, plagiarism and upvoting and review cabals, Amazon either chooses not to invest or not to invest enough to prevent all of that bad behavior.

IMHO investing 1/2 of 1% of the KU pool is enough to fund human review via their Mechanical Turk program. It wouldn't be perfect but it would eliminate most of the fraud.

Re: Some Amazon Reviews Are Too Good to Be Believed – They're Paid For

#49
post #20

It's common knowledge among most techies that the majority of Amazon reviews are fake and that counterfeit items are a big issue, but most of the general population is completely unaware. I was searching for an inexpensive pair of polarized sunglasses last week and the "Amazon Recommended" ones had an F on FakeSpot with 70%+ fake reviews.

You'd think that Amazon would have the machine learning resources to detect likely-fraudulent items with high accuracy, should they be interested in preventing such obvious fraud.

They have the resources. What they lack is the proper incentive. As long as fake reviews don't completely destroy the customer experience to the point that people forego any other advantages (mostly convenience) of shopping there, Amazon will continue to reap the short-term benefit of more apparent variety, more apparent quality, and more sales. Any effort beyond keeping the fakery just short of destroying the brand isn't worth it to them.

Re: Some Amazon Reviews Are Too Good to Be Believed – They're Paid For

#50
post #26
post #23

There was a great Planet Money podcast on the genesis of ReviewMeta which I've been using on Amazon reviews since. https://www.npr.org/templates/transcript/transcript.php?stor...

I've been using https://www.fakespot.com/ to check review for fakes on Amazon.

And I've started using BOTH reviewmeta and fakespot. Good results so far.
Post reply on HN