Re:Advertising Without Signal: The Rise of the Grifter Equil...
by godelski on Sunday July 20, 2025 @12:32AM
(#44621998)
They can definitely solve some problems. There's a number of items that are well known to have common forgeries. Weirdly common is things like Lamy pens[0]. I've gotten legit ones from amazon but seem fake ones.
One big problem I see is if you go to the storefront of some seller there isn't a way to see all their items. I find this baffling, and a clear red flag. Just start clicking on the stores when you're poking around next time. It's very common for things like pet toy sellers or even electronic component retailers. I've seen even certain colors of products be listed on a different page and sold for much cheaper.
There's also the classic example where a page was originally for one product but now has another. You look at the reviews and they're talking about a different product, sometimes related sometimes drastically different. This is actually a great use case for some ML. Flag products if a classifier says the item is significantly different than the previous photos. Could use some basic NLP to check product titles and see if they match. It creates some issues but Amazon has such a big problem that it would be worth it for the customer experience.
But I think the problem is all of this doesn't "make more profits." If it's hard to verify a product a user will spend more time on Amazon. If the idea is that more time increases likelihood of buying more things then a little fraud is good for them. Worse for the customer but better for them. I'm not sure how you solve this problem
One big problem I see is if you go to the storefront of some seller there isn't a way to see all their items. I find this baffling, and a clear red flag. Just start clicking on the stores when you're poking around next time. It's very common for things like pet toy sellers or even electronic component retailers. I've seen even certain colors of products be listed on a different page and sold for much cheaper.
There's also the classic example where a page was originally for one product but now has another. You look at the reviews and they're talking about a different product, sometimes related sometimes drastically different. This is actually a great use case for some ML. Flag products if a classifier says the item is significantly different than the previous photos. Could use some basic NLP to check product titles and see if they match. It creates some issues but Amazon has such a big problem that it would be worth it for the customer experience.
But I think the problem is all of this doesn't "make more profits." If it's hard to verify a product a user will spend more time on Amazon. If the idea is that more time increases likelihood of buying more things then a little fraud is good for them. Worse for the customer but better for them. I'm not sure how you solve this problem
[0] https://www.reddit.com/r/fountainpens/comments/q14p5w/are_la...