Hacker News Reader

> An alternative explanation for this is that Amazon is struggling with identifying fake reviews

I seriously doubt this.

I review most products I purchase and most of the time I leave a description and written review. I tried to leave a 1 star review on an Anker Auxilary Cable I bought that was an "Amazon's Choice" product. I tried for 3 days and every time I receive an error trying to review the product telling me reviews were unavailable. I then tried giving it a 5 star review and it was immediately accepted, at which point I immediately changed the review to 1 star and explained why.

* So oddly I just tried to review a piece of electrical safety equipment I purchased through Amazon a few months back. I tried to give it a 3 Star (and 5 star) review and received the "Sorry, we are unable to accept your review of this product." notification.

Oddly this product has zero reviews. Are sellers allowed to disable reviews for their product listings? If so, do the existing reviews remain visible?

If your maybes are valid it simply means that these Amazon guys are new in the business and will figure out how to design user interface once they have the resources.Lets give them a slack, aye?

Seriously though, I canceled my prime after I got a fake SD Card that was sold by Amazon. There was no straightforward way of returning it, had to contact customer support.

Later I looked closely and I noticed the reviews warning against fake SD Cards but these were buried in enthusiastic 5 star comments celebrating their new SD Cards and trying to tell the world what a good purchase that was.

Maybe it’s just due to the scale, maybe it’s in the business model but dealing with fakes wasn’t working for me. If I wanted the eBay experience I would have been buying my SD Cards from eBay.

It's not necessarily a resource problem, from my time there there is an institutional lack of respect toward UX and front end and this is might still be the case.

There's also a myriad of custom Perl templates that are essentially impossible to maintain. I once had to investigate where a feedback form sent feedback to, the answer was nowhere for about a year.

When were you there?

You are throwing together different issues.

Amazon has lax standards as to who can use them to sell their wares (leading to fakes being sold). That’s not UX or a UX dark pattern, that’s a business strategy. (Which doesn’t make it necessarily better or worse than if it were a UX dark pattern if you are judging it morally, it’s just something different.)

Amazon makes it hard to initiate returns. That might be a dark pattern. (In my experience the returns process for things Amazon sells directly or more or less directly, returns are straightforward and simple. Third party sellers are here, again, the problem. And that’s then more a mix of UX and business strategy, where third parties either cannot or are not required to use Amazon’s returns process but have to provide their own, e.g. fall back to writing emails back and forth.)

Reviews are overly enthusiastic, even though the sold product is not genuine. This might be an issue with fake reviews or uncritical customers (themselves unable to recognize fakes) or both – but it’s not really a UI dark pattern.

I would also argue that Amazon’s way of surfacing negative reviews (with the stars histogram and ways to filter reviews by number of stars) doesn’t suggest that it’s Amazon’s intention to hide negative reviews …

O.K. Look, when a website tells me that I'm purchasing a "32Gb Kingston UHS SD Card" by "Kingston" I expect to receive a 32Gb Kingston UHS SD Card.

There are details like "Sold by SomeScammySeller" or "Fulfilled by Amazon EU S.à.r.l" on the page and there are caveats like showing the same reviews for the same item sold by different sellers.

After the title, there would be a clickable link with the brand name after the title making the impression that you are buying something from that brand.

It Makes it very hard to know WHAT and from WHOM you are buying from and what are all these reviews about. Sometimes you change colour and everything changes.

You need to study the title, you need to study the comments, you need to study the Amazon's way of doing business and have understood what Fulfilled means, What is Amazon EU S.à.r.l, What is their relationship with the sellers and what guarantees they provide.

I don't know how dark or light coloured are these design choices but, in my personal opinion, these create a lot of cognitive loads and combined with not providing a return option and requiring to go by the customer support looks sketchy to me.

What companies need to start doing is digitally signing their products (e.g. put a unique code on each box that is generated from a private key). Then Amazon and/or sellers can go online and verify this key to ensure their wares are not replica.

I'm curious how you think that would work. They could sign the serial number, I suppose, but I don't see why the fakes wouldn't just recycle serial numbers at that point. For fast moving parts, they could include a date, but you could still just copy some other one from recent checks.

I think how it would work (if I amazon wanted it to) is that the supplier would send the list of signatures for the next 200 products to amazon (which would then check for duplicates). Presumably these signatures could be scanned from a barcode format programatically.

Also, consumers believing they had a fake product could then go to the website of these manufacturers, who would have an incentive to make a web-ui to reveal duplicates/phonies.

I'm not gonna pretend I've invented a perfect end-to-end solution here, but if I'm getting 90% of the way there in 10 minutes then let's not pretend it's unsolvable.

Apologies if my tone made it sound dismissive. I'm genuinely interested in how this could be made to work.

I agree in thinking a full on handshake would be required. Not sure how instrumented all of the suppliers would be. And I'd imagine that is where a lot of the pain to build a solution like this would be.

Any other approaches that you could think of that don't require cooperation from every party?

There is also the question of cost and basic reach. You assert this gets 90% of the way there. Do we really think the fake problem is over 10% of traffic today? I'd be surprised if that is the case. I don't have numbers, but I would expect this to be in the .1% range. At the scale of a large store, that is still a giant number.

Put the serial number under a scratch off. Once the serial number is registered by a customer via an online database, future verification would show other purchases as counterfeit.

It has some drawbacks:

- You need a significant number of people to start registering their purchases online, which no one does

- I've heard counterfeiters sometimes operate from the same factory. Legit serials could be generated for fake parts.

- Customers will obliviously try to pursue returns with the victim manufacturer instead of Amazon.

>I've heard counterfeiters sometimes operate from the same factory. Legit serials could be generated for fake parts.

A few years ago I ordered 60 computers from Dell with office 2016. About half of the 2016 serial cards were from the US, they all worked. Of the other half from China, half were already registered under office accounts. These were cards with scratch off covering part of the serial.

What? They make returns very easy to do. There's a form that pops up and many choices for reasons.

There was no return option on the SD Card, it was marked as "not eligible for return" so I had to deal with the customer support.

I actually went ahead and dug out the conversation for you.

This is my refund request: https://imgur.com/bthMMoN This is the response from Amazon: https://imgur.com/xWK8FVH

Yeah, they refunded me without a fuss but I prefer an experience where the items arrive in expected quality and I keep them.

I think I actually kept the Prime membership for a while but I found myself disillusioned and felt uneasy when shopping at Amazon and later canceled as I was no longer buying stuff from Amazon.

So all it took was an email, they didn't even require you to return the product before refunding you, and you're unhappy with that for some reason?

Absofuckinglutely. Anytime you end up with a fraudulent counterfeit it's a hugely unhappy situation for a number of reasons. The biggest for me would be: What if I hadn't caught it? Maybe I'm using some inferior counterfeits right now that I don't know about that Amazon ripped me off on, that might fail catastrophically. And also, there's plenty of times where you need a product by a specific time, and not having it at that time because you wasted a whole shipping cycle on a counterfeit is painful.

When you end up with a counterfeit, you are being defrauded. How could anyone be happy in that situation just because your money was refunded? I just wanted what I was trying to buy. I didn't want free junk that wastes my time.

I spent all day thinking that my raspberry pi zero was faulty, trying to figure out why it’s not booting(FYI, very hard to obtain).

Yes, I’m unhappy. A happy purchase is not one where you get a refund for low quality product, a happy purchase is a purchase that you end up spending your money on a product that satisfies you.

I don’t care about 7gbp, that’s not the issue.

Returns through Amazon have become increasingly difficult and spiteful in the past two or three of 20 years I've been using them.

Last week I returned a rotary airer that was damaged upon delivery. Amazon online 'help' told me I had to find a courier that could take the outsized parcel AND I had to pay out of pocket for it. They insisted that it be returned. They didn't refund the full postage cost or the cost of packaging.

Amazon are throwing away their hard-won reputation by penny-pinching and treating customers as badly as other retailers. I'm out.

Yeah, I mean, he got a refund and didn't even have to return anything. How dark is that pattern?

Having said that, Amazon's product mixing is causing a lot of problems and trust issues for a lot of product types. I really hope they fix it.

Me not getting a refund would have been a fraud. I would have filed a chargeback against that fraudulent purchase.

A dark pattern is not about taking someone's money an running away but it's about impacting their decision making by designing processes or interactions that exploit common behaviour.

An example of a well known dark pattern is adding things to the shopping cart just before check out. Another one is starting a subscription with a click of a button but requiring documents to be faxed for the cancellation.

On my specific case with Amazon, the dark pattern is to make it hard for me to spot a fraudulent item and not making the standard refund process available therefore increasing the chances of purchase and decreasing the chances of refund by making it a hassle. Well, maybe it was not intentional, maybe it was a screw-up but that's again a lot of maybes.

Go do some digging on Amazon and look at all of the obviously fake SD cards out there. Then ask yourself, is this helping Amazon's business? This is a problem, and they treat it as such, they simply haven't found a suitable solution beyond simply making refunds with almost no questions asked. Unfortunately for their business, this actually enables a different kind of fraud that is also a real problem for them.

That is vastly different from dark patterns about trying to inflate review scores or encouraging subscriptions to unwanted services. They do those things too, and pretty effectively.

Any sources on your claim that they are actively trying to solve that problem and that they are failing at it?

https://www.bloomberg.com/news/articles/2016-11-28/amazon-ge... among others.

I don't think they can really solve it without a complete top-to-bottom shift in how they track and inventory products internally. Most of the fixes that I've seen proposed just lead to the vendors and merchants gaming the system in different ways.

Of course they will say that they are cracking down on fakes.

The only thing Amazon is struggling at is prying open their checkbook to pay for oversight and accountability because it would eat into their phenomenal profits. AWS alone generates $20+ million net profit a day (while we perpetuate the rumor it saves money LOL). It wouldn't cost very much to solve all of these problems with human labor, but not solving problems is incredibly important to companies too just like not paying taxes and not having toilet breaks for your warehouse workers.

One situation I found where AWS saves money is in over-capacity situations. If the a company builds a data center that is too large, they are incurring costs for under-utilization. Paying a 50% premium on 100% of resources might still be better than paying 0% premium on 200% of resources. The cloud versus on-premise discussion has many variables, so it all depends on the situation.

Amazon has been removing reviews and locking listings from receiving new reviews too while they investigate the past couple of weeks. This is probably the reason.

I've also heard that if a listing receives too many reviews in one day, they will prevent it from receiving any new reviews. I think certain unscrupulous sellers have also figured out how to take advantage of that, which is a little funny. Just post all your fake reviews before your unhappy real American customers wake up...

Well, this bad UX has a positive consequence for Amazon, unintended or not.

Products with good reviews generate more sales than those with comments stating problems with the product or delivery process, specially among infrequent buyers. (Frequent buyers may learn to ignore comments altogether, or at least spot valuable comments from fakes).

So the bad pattern is not allowing fake reviews as much as it is rejecting legitimate negative reviews.

I would, however, postulate that Amazon is aware that fake positive reviews only buy them short term benefits with long term disadvantages.

Having a cheap/free way to evaluate products by letting the customers do it is an advantage, especially if your goal is to sell everything (so you have no hope to evaluate it all yourself) – but it’s only an advantage if some useful signal is provided.

With strings of negative purchasing experiences customers might begin to look somewhere else. That’s why Amazon has a vested interest to provide reviews that are more signal than noise.

Except, will they look elsewhere? The current trend seems to suggest otherwise.

Also, will the sporadic negative experiences be enough for users to leave the site, if they are distributed among all the infrequent buyers?

Maybe not leave the site, but at least distrust negative reviews. I recently bought some towels that had amazing reviews. The towels sucked. That won't make me stop using Amazon for things in which I already have confidence but it will stop me from relying on their reviews. I will buy things like towels elsewhere.

This episode of reply all show how hard it is to identify fake reviews and what seller are doing to fake it. Really interesting in my opinion. https://www.gimletmedia.com/reply-all/124#episode-player

I actually didn't like that episode because she very clearly said she spent no time looking at the product and just clicked buy. But there are legit reasons to want a UK plug in the US (maybe you're traveling and want to have it with you when you leave). The fact that the seller is trying to game the rankings is more an issue with the rankings system than with the seller.

Gaming the rankings/listings isn't an issue exclusive to amazon, every single website has this issue, in fact even phone books (remember those) had to contend with people listing their name with leading A's to try to get to the front. Gaming listings is as old as lists and until we can come up with a solution to that problem we're always gonna have problems like this.

If the person who wrote the post could immediately identify the clearly fake reviews I see no reason why Amazon could not use the same method.

I did a lot of work on anti-botting some years ago. It's not as easy as it looks. Often you spot a pattern that is obviously fake, like here, but discover it can't generalise without huge numbers of false positives.

For example I would expect lots of legitimate users to review the same small set of TV shows or video games if they buy into a series or franchise.

FakeSpot and ReviewMeta seem to do a decent job of identifying obviously fake Amazon reviews.

I suspect that Amazon just have an incentives problem - purging their site of fake five-star reviews would go a long way to restoring trust in their platform, but it would negatively impact sales in the short-term. From what I've heard, Amazon has a very decentralised and data-driven management culture, which is antithetical to the short-term pain/long-term gain implicit in fixing their reviews problem.

How do you know what their FP rate is?

But equally - does it matter?

I would rather have less, but more accurate, reviews than have all of them but many of them being fake.

It seems like a use-case where a high false positive rate would be acceptable. Unlike say, medical diagnosis or something.

The problem is that very few really "good" reviews actually get produced. If you accidentally filter out the really thorough and useful ones, you can end up with "real" reviews that aren't well written and don't really focus on actual product issues.

Take, for example, the review of the weights in the article. I would have ignored his review, even if it had been posted. He reviewed Amazon's warehouse rather than the product on hand.

If legit reviewers have their work deleted because of FPs they get annoyed and won't write more reviews, they may even stop using the site entirely.

Deletion is not the only option. They can weight the contribution of reviews to the overall score. They can weight the display order of reviews on the product page. They can weight the display order of products in SERPs. Assigning reviews, reviewers and sellers a "fakeness probability" allows you to do all sorts of subtle, useful stuff that clearly isn't being done right now.

They could shadowban the reviews - so the reviewer would think it was being displayed.

A bit deceitful but maybe better for UX.

If I want to ignore fake reviews because they don't provide me with useful information is it unreasonable to think I would also find reviews that have the same characteristics as a fake review unhelpful? I don't want to spend all day looking through reviews so I think the false positive rate isn't as big a problem for this application as some others.

I would expect that the reviews which are a result of purchase would be treated differently? Especially if not returned, or if the customer is returning less than X/Y products? Amazon has the advantage that their users are known (to them), so it's a lot easier to fight fakes.

As in employ someone to go and trace users who post reviews to see if they’ve reviewed the same products?

Maybe you meant make a system that can do it, and do it without false positives to not upset legitimate users when their own reviews are removed.

I’m confused about the ease people expect this problem to be solved. It’s so easy to solve it exists on every popular platform: see the paid, fake reviews on the Apple App Store.

are you kidding me? the algorithm for "top" reviews is approximately synonymous with "hide negative reviews". a stretch to see how it benefits amazon? of course reviews influence purchasing, that's why they're there

I cannot buy any argument along the lines of "maybe it's just bad UX." Amazon has the resources to study and hire the best UX people in the business. There is no excuse for them having a UX problem except that they don't want good UX.