Hacker News Reader

Comment by voidwtf | Hacker News Reader

This is among greatest abuse Amazon allows in its product listings. Any change to a page should reset the reviews unless otherwise vetted by an Amazon employee. Changes shouldn't be material in nature, only corrections. A new revision/version of a product IS NOT THE SAME PRODUCT. Different product == different listing.

View on HN

That would be an easy way to get rid of negative reviews.

There already is an easy way to get rid of negative reviews. Just create a new product. That’s what they do.

That might be going a little far. Adding an extra-large size to a listing for a T-shirt (for example) shouldn't invalidate all of the existing reviews.

A more targeted approach might be to disallow sellers from adding variants to products which are not currently for sale. I suspect that would cut down on most of the abuse.

fair point, that said it would take amazon a blip of resources to have some NLP diff to delay validation of title changes that are too strange

> Adding an extra-large size to a listing for a T-shirt (for example) shouldn’t invalidate all of the existing reviews.

If you want to deal with this problem rather than channel more of it into a place some of it already goes, it should, because completely different products are added as “sizes” or “colors” of existing products already.

If you add a review fraud control and leave this an escape hatch, well, the results are predictable.

Adding something should trigger a manual review by an amazon employee.

An instant ban for vendors detected and confirmed to be doing this. Amazon has plenty of ability to deal with this, from a ML detection system to human validation. They just don't want to.

They do, but they just pop up as a different seller under a different name with a different business address and the cycle goes on. That's why all the sellers have weird names and make no effort to drive any sort of brand awareness or loyalty.

All that tech, platforms, manpower and literally all the backend data about it (access patterns, descriptions, sales, back accounts, user agents, IPs, everything) and they can't detect something that's blindingly obvious to anyone paying attention?

Even more than that, they require government IDs and other documentation to set up seller accounts, but at the end of the day it's easy to come up with new straw front people for your scam, with clean papers, addresses, and bank accounts.

Noone at Amazon is looking at the id. If they did it would add 'friction'

So mark a vendor as "new vendor", add mandatory ghost shopping to all product skews (postage returns paid by vendor), funds only clear after product arrives + n days for new accouts, etc to build in security that products sold are legitimate. All you need to do is make cost of standing up a scam more expensive than the real thing.

The weird brand names are primarily to make obtaining trademarks easier (though I'm sure easily changing business names after getting busted for fraud is a nice bonus): https://www.nytimes.com/2020/02/11/style/amazon-trademark-co...

That's something different though somewhat similar.

Why would they not want to? This is a serious business problem for Amazon, isn't it?

I don't know why they don't want to, but a database query for "\d{2}TB SSD" with a price under $100 is surely not beyond a global top-three AI company with a half-trillion dollars of revenue.

Reply on news.ycombinator.com