Algorithm hunt
- Will Emery
- Apr 30, 2024
- 2 min read
Blogpost #3
Research question: How sensitive is the Amazon recommended algorithm and how easily can it be overrun with a certain type of product?
Steps:
1. Make a new Amazon account in an Incognito tab so that nothing from my personal browsing history influences it.
2. Begin by searching for certain products and clicking through to the pages. For my run, I started off by looking at exclusively tech-related products. Some examples of categories that fall under this broader category are headphones, laptops, watches, and phones.
a. This screenshot is the result of my initial run with tech products.
3. I next looked up 3 books to see if my recommended feed would shift at all.
a. This is the result of my experiment. As you can see, only one book only showed up in my "keep browsing for" section. So all in all, this was showing that the Amazon-recommended feed is maybe not as sensitive as I may have initially believed.
I chose to examine this algorithm because, in my experience, this algorithm has been very sensitive to when I will look at a single product in a category. This may be because I have more regular browsing habits on my own account as opposed to the browsing habits that I exhibited on this account. The high concentration of looking at probably 30 products in a given category and then the sudden shift to looking at a few products of another type wasn’t enough for the algorithm to think that I wanted to see more of category #2.
The main reason for my finding is probably due to the product tags that Amazon uses on the backend. Most of the things that I was looking at could have had tags similar to “mobile technology” so it was only recommending products with those tags. I hadn’t looked at enough items with the tag “book” so I figured that this was lower on my priority thus I would be less likely to buy it. The way that this could translate over to the world of circulating my online portfolio is by tagging my work with a wide variety of key terms to hopefully infiltrate spaces where it normally wouldn’t show up.
Interesting experiment! One of the things I've noticed browsing Amazon is that it tends to get attached to things you fell down a weird rabbit hole you fell down months ago, and doesn't instantly adjust like other sites, so it's cool to get experimental confirmation! As for the application, that's a great strategy idea. There are so many ways to get attention online, and I think to really get a portfolio out there it's important to be using all of them in a scientific way.
Hey Will,
I also noticed that key terms are really important when trying to get my portfolio to the top of the search results. I'm not entirely sure if Wix has a hidden "tags" page to do this, but if there was it would certainly be helpful. I've always thought it was interesting how much more specific and sensitive algorithms get even after a few searches. It makes me wonder, when is the search the recommended one due to others searching similar items, and when is it actually directly advertising a product to you? -Brooks Gilbert