Amazon A9 Patent

A9’s patents and how they work

All Amazon sellers are affected, but many don’t even know about them. Let’s talk about A9’s patents and how they work in Amazon’s algorithm.

It’s a bit of a technical post, but it’s worth doing this exercise to understand how and why our products appear in a certain position in the marketplace’s internal searches.

I’m Jordi Ordonez, and I’m an Amazon Seller and consultant based in Barcelona, Spain. My background is in SEO and I work both for Vendors and Sellers.

What is Amazon A9?

Those who did not know it previously, with the introduction will have already got an idea. Basically, it is the algorithm that, based on certain criteria, determines the relevance of a certain product for a search performed in the Amazon search engine.

This set of rules analyzes the entire database they have indexed, and sorts the organic results.

Doesn’t it remind you of how Google works? Normal, since originally and before moving to an internal use, it was conceived as a competitor of the Mountain View search engine.

Its extraction, indexing and ranking process works the same as Google’s since 2003. This is explained in a patent called Search query processing to provide category-ranked presentation of search results.

Amazon A9

What are patents?

This type of technology is also governed by law, search engine development teams devote hundreds of hours, talent and resources to defining new ways of interacting, sorting, categorizing, hierarchizing…

So, when they come up with a process that can make their tool more accurate than the competition’s or simply improve the user experience by presenting more relevant results, they proceed to register it to restrict its use by third parties.

How A9 patents work

New ones are made every year and each one works in a certain way and powers a certain area. Anyway, there are things they have in common.

Basically, what is recorded is a flow chart that visually details the entire process step by step. Looking at one of these patents is striking because of how schematic they are. It could be the operation of a combustion engine or a coffee maker, but it is software.

But the best way to see how it works is with some concrete examples.

#1 – Indexing and presenting content using latent interests

The name is very descriptive of what this patent does. What it seeks is to generate a relationship between what the user is interested in and the products it shows. In other words, it monitors the content that generates interest in the user outside the Amazon environment.

Thanks to the information extracted from those external contents and reviews, added to the internal ones, the algorithm establishes the relationship between those data and the ASINs of its database to, in this way, provide more relevant answers for those undeclared interests within the Marketplace ecosystem.

This is a recently registered patent, specifically in July 2021 and you can check it out here.

Amazon A9 Search Engine

#2 – Increases in sales rank as a measure of interest

What is Amazon looking for? Yup: to sell. Therefore, it makes perfect sense that an increase in sales is interpreted as a positive sign of relevance. If something sells a lot, there is a lot of interest, so it justifies showing more.

This patent defines exactly what you’re thinking in: the Best Seller Rank or BSR.

This is what determines this patent. It reviews the sales ranking of each category every hour and, based on this acceleration of the rhythm, it orders the positioning of the products.

Important: this does not mean that the organic keyword rankings come out of the Best Seller Rank; this is only true for the categories. It even changes the average Best Seller Rank from category to subcategory.

The patent for Increases in sales rank as a measure of interest, was filed in 2016 and it’s available here.

#3 – Machine learning based data base query retrieval.

Since the time when science has advanced enough for machines to be able to learn from our behaviors, search algorithms have made a substantial leap in efficiency.

Amazon introduced this patent in A9 that uses historical information of products similar to ours, to perform a classification of that new product for which it does not yet have a reference.

This has been internally dubbed Cold Start Service (aka Honey Moon Period) and is intended to prevent a new product from being weighed down with low rankings from the start.

In case you are wondering, as we already discussed, it is also related to that grace period called “Honey Moon Period” in which they increase the visibility of newcomers in a somewhat artificial way, to get them to start generating information with which to evaluate it definitively.

This patent was registered by Amazon on March 8, 2022. Here is the link.

#4 – Providing location-based search information

Another important aspect that has an impact on searches is the geographic location of the user interacting with the search engine.

With the patent we are analyzing, A9 is able to produce different rankings based on this variable. In other words, customers from two different locations are most likely to see different rankings for the same product.

Here you can see the documentation of this patent dated June 13, 2017.

This is just a sample of A9’s patents and how they work, but if you are interested and keep pulling the thread, you can get a very clear view of how Amazon’s search engine behaves. Here’s how to go down the rabbit hole:

You May Also Like

About the Author: Garlic Press Seller

Leave a Reply

Your email address will not be published. Required fields are marked *