this post was submitted on 15 Dec 2024
270 points (96.9% liked)
WTF
4592 readers
2 users here now
founded 2 years ago
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Amazon uses an offline item to item comparison table. Every so often it will collect all items viewed or bought by people who bought this product to make a table of what to recommend. If an item is added right before this information is pulled, it's likely to have a small sample size. This will then remain til the next update. So it's possible this was based off only a handful of people making the results very skewed and that it hasn't updated since.
Source: my professor who helped design Amazon's algorithm
I hate to be that guy, but who is your prof?
I work at Amazon, and my understanding of how the teams and services that put together this kind of functionality for recommendations is that they'll be updating this and testing against MANY variants in a given time to optimise where possible.
I'm not aiming to call you out, mainly curious, because right now unless your professor is an active employee at Amazon or a part of the scholar program the most this could be true is that they created one variant that was once used.
I won't say their name because that would indirectly identify me but they left Amazon to teach a few years ago so this information could absolutely be outdated.