Steam handles that problem by showing two scores for the item (games in their case): the total score, and the "past 30days" score. This way you can see if something is going wrong recently.

I love that mechanic. I think it's a lot more in the forefront with things like software that gets auto-updated, where the publisher pushes an unfortunate patch, or fails to follow through on a promise which was previously given the benefit of the doubt. I don't think physical products have yet caught up to the idea that this notion is relevant to them as well.

The especially difficult part when it comes to reviews for physical products in that it just isn't sexy to benchmark an old piece of hardware. What makes the situation even worse is that most reviews aren't centralized (that is, you google "productname reviews") and so to a large extent the visibility goes to whoever publishes first. Your brand new benchmark announcement gets clobbered by the pagerank of the reviews that happened right when everyone was interested in the product.

I had long assumed that Valve added the "past 30 days" aggregate to show when games had improved rather than gotten worse. (i.e. early access games) Kind of interesting that it helps clarify changes in both ways, and implies very different incentives for the game publishers/sellers.

It's a poor argument and they are misleading their customers. You don't buy on the specification you buy on the reputation. If $creditable_reviewer says it's one of the best drives they've ever used, you are going to go for it.

When you buy the exact same product and end up with something that has substantially lower quality parts than what the reviewer had, you've been defrauded.

Edit: it hurts the reviewer too, as I now have reasonable evidence that $review_body is unreliable. If the fraud never comes to light, I may stop using they publisher's reviews.

Poor in what sense?

People were certainly mad about it, but the companies didn't have to face a lawsuit, so as far as they were concerned it seems like it was a great argument :-\

> When you buy the exact same product and end up with something that has substantially lower quality parts than what the reviewer had, you've been defrauded.

How exactly do you define "substantially"? No matter where you draw the line, you'll get people who will toe it and that will be legal. And in any case it's very hard to argue fraud when the company itself didn't tell you that the device was that good. You'd just get companies (rightly) saying that they aren't responsible for purchasing decisions you made based on things that third parties have said.

From someone like Amazon's perspective, it's also not reasonable to declare that absolutely no changes can be made to a product to keep it under the same listing, which is really the only way you can make this a black-and-white issue. It would also be doing consumers a disservice if every time a trivial thing changed (e.g. logo placement on the box, or the manufacturer adds an extra screw to make the body stronger, or whatever) the entire history of reviews and ratings got reset. So the best solution we have right now is to say "look man, don't be an ass"... and unsurprisingly, people who want to be an ass are more than happy to disregard this in order to make an undeserved buck.

There's another solution, which is that reviewers who find that their good review is being used to flog a substantially worse product (and in this case, the reviewer themselves gets to determine what is substantial) disown/update that review and/or refuse to review that manufacturer's products in the future.

that would require a lot of thought and tracking on reviewers part that I'm not sure most reviewers would care to do. Esp since it's for the benefit of others (amazon/manuf) and not themselves. If you pay them then you throw off the balance.

Also, it would be too late!

From a few things I've read, I don't think this is specific to Kingston. It's apparently pretty common for offshore factories to optimize for cost while keeping the product apparently the same.

There is a principal-agent problem. The company whose brand is on the line needs to be diligent about catching it in QA.

I seem to remember reading somewhere that Walmart will ask the manufacturer to build a special lower-quality version of the same item to keep costs down. So a direct price comparison can be misleading.

Yeah, I wonder if it's always the fault of the brand, actually.

I remember that during the FTDI fiascos[0] one of the often-raised complaint is that many vendors don't even know they're using fake FTDI chips, because it's the factory that decides to replace the genuine chip with a counterfeit to cut down on costs.

Then again, a cynic in me sees this as a plausible deniability scenario - the brand can always say that they didn't know about the component switchup, "it's that Chinese factory that did it, and we're investigating!".

[0] - FTDI supplies the most popular USB-to-serial chips, which are present in pretty much any USB-connected device you have; on two instances they released driver updates that detected and bricked devices with counterfeit USB-to-serial chips.

There's a fine line here between "counterfeit" and "clone". It's only a counterfeit if it's sold as a genuine FTDI part.

Gomorrah ( https://www.amazon.com/dp/0312427794/ ) has an interesting discussion of the workings of the fashion industry in Italy. It goes like so:

1. A fashion label invites a bunch of clothing manufacturers to a pricing auction. They all meet somewhere.

2. The representative of the label describes the clothing they want. The manufacturers bid for the contract, with bids taking the form "X many items within Y period of time at Z cost". For example, 800 items / 20 euros per item / 1 month.

3. Eventually, the brand will accept a bid. At that point, every manufacturer may commit to fulfilling it. Any manufacturer accepting the winning bid is given the material for free.

4. Whoever delivers first gets paid. Everyone else is left with a partially-complete order that they can no longer sell to the brand that commissioned it.

(4a. Manufacturers who show a history of accepting bids for the free fabric and then failing to deliver get blacklisted.)

5a. The brand sews its label onto the clothes it bought and sells them at "high" prices.

5b. The slower manufacturers may sell their too-slow clothes to the mafia, which sews a counterfeit label onto them and sells them at "low" prices.

So now we have the mafia selling "counterfeit" clothing that would have been "real", if only it had taken a little less time to create. The fabric and cut are to brand specifications because the clothes were made at the brand's direction to bear their fancy label. Would you call those clothes counterfeit?

Bose totally did this with their MIE line of headphones. I bought a pair right when they were introduced. They seemed very well made and I loved them. Sadly, that pair got stolen. The next pair was the same but... different. Several pieces just seemed flimsier, cheaper. The cord was a bit thinner, the plastic remote was creaky, earbuds felt were less substantial, etc. Melted that pair on a radiator, unfortunately. On to pair #3. Everything feels cheap, which is annoying when you spend $120 on a pair of earbuds. Cord leading from the remote to the earpiece is approximately the thickness of dental floss and breaks inside of 6 months. No more Bose for me.

This gets at an important point in the Lean Manufacturing philosophy. The typical business chases profits and works to reduce costs. But Lean businesses focus on delivering value while reducing waste. For your headphones, somebody was cutting costs without regard to value vs waste.

I think one of the reasons this is so common that anybody can cut costs. ("Your budget next year is 10% lower. Make it work!") But it takes thoughtful experts to discover and eliminate waste.

> It's apparently pretty common for offshore factories to optimize for cost while keeping the product apparently the same.

It's called "value engineering" and is commonplace in every business. It's like taking working code, and then optimizing it over time. Finding ways to cut costs goes on all the time, and it does not necessarily imply cutting quality.

If they are caught they could:

* Drop the product from the store completely.

* Put a Warning Misleading Listing stamp on the listing.

* Penalize the seller in the results.

Add a negative "age" score that slowly increases over time, like how reddit articles are ranked. This will bias the average ratings towards newer reviews. This can also be done in the algorithm that decides which reviews to showcase.