July 2025
Joe Solari
So, I decided to reverse engineer Amazon's algorithm.
There is a lot of myths and hearsay floating around about how Amazon really works an my hope is that by doing a deep dive I could have better answers.
You know what I'm talking about.
Those forum posts claiming you need to do X, Y, and Z to get visibility, but nobody can actually explain why those things work, or where they learned this information. Sure, you might see a screenshot of results occasionally, but there's rarely any underlying data. Then, three months later, the same person is posting in the group asking if something has changed because they're no longer seeing those killer results.
I figured there had to be a better way to understand this stuff, so I went straight to the source materials – research papers, patent filings, academic studies.
As you read on, understand this isn't about you simply taking ideas and applying them. Nor is this some ploy to get you to buy something. I did this to better understand the game I choose to play. If you're really serious about publishing, this stuff matters, and you need to develop your own ideas and strategies, not just mimic others.
What I discovered was fascinating: the world of recommendation systems is actually quite small. The same names keep popping up across different papers and patents. These researchers move between a handful of top universities and tech companies, building on each other's work and citing each other constantly.
Once you start mapping out this network, you can trace the evolution of ideas from academic papers to real-world systems. The person who published that breakthrough paper at Stanford in 2015? They're now a principal scientist at Amazon. The team that solved collaborative filtering at Netflix? Half of them ended up at Google, the other half at Amazon.
It's like following a trail of breadcrumbs through the academic and tech world. And when you understand the intellectual lineage behind these systems, you start to see the logic in how the recommendation and ranking actually work – not just the speculation and guesswork that dominates most discussions about Amazon's algorithm.
Before we dive in, let me be completely transparent: I haven't talked to anyone at Amazon about this, and I don't have any insider information. Everything I'm sharing comes from publicly available research. I just took the time to read it and write about it.
By understanding the people and ideas behind these systems, I think we can get a much clearer picture of what's really happening under the hood.
My work here is in two parts;
You may be inclined to just jump to part two, that’s ok but you’ll likely be driven back to part one for better understanding of the why.
Right after this introduction will be the Too Long Didn’t Read section with some action items. Ninety percent of authors will just read that and go implement.
That’s cool. I did that to satisfy those who just seek actionable items as well as signal search crawlers. Those five hundred or so words make up less than ten percent of what is written here and less than one percent if you include the research hyperlinked to this page.