Ever wondered how when browsing through the Amazon website/app there is always a list of products that are “Recommended for you…”? Do you often sit there and think “How on earth do they know that I might want to buy those unicorn socks with matching scarf….”? Well, it’s not witchcraft, Blackmagic or lucky guesses. For the past decade or so Amazon has been spending a lot of time investing in their machine learning algorithm that powers their recommendations service on their sites. Its getting that good at recommending products, that a whopping 30% of Amazon.com’s page views are from recommendations alone!!! That is an enormous amount of traffic.

So what is Amazon Personalize?

Amazon Personalize is like having your very own Amazon.com machine learning recommendation system, running for you 24 hours a day. Once data has been supplied to Personalize, it gets to working learning from that data, to start predicting what products your customers may be interested in purchasing.

Benefits of using Personalize
  • AWS managed service
  • Out the box ML
  • High quality recommendations
  • Improve user engagement and conversions
  • Get started in a couple clicks
Some use cases
  • Product recommendations - Product recommendations based on prior user behavior, preferences, and shopping history
  • Similar item recommendations - Suggests similar products to those you have viewed in the past
  • Personalised rankings - Promote specific content to users based on personal or busniess needs
What makes up Amazon Personalize

Personalise is made up of main 4 sections:

Datasets Event Ingestion Solutions Campaigns
These is a file or files in CSV format of historical data to import. It’s a requirement to get started. This allows users to create real time ingestion of data from your applications. This is where the “magic” happens. Solutions are where you train Personalise using machine learning models (recipes), using the data supplied. A campaign exposes the results of your solutions back to your applications via an API. It can also provide some analytics of the solution usage.
End

So what do I think about Amazon Personalize??? I think its a great piece of tech. For someone who is slightly interested in Machine Learning, but doesn’t have the time to learn it in depth, this service allows people to quickly and easily setup a recommendations engine, and all the heavy lifting is handled by AWS. As usual you get all the AWS goodness of fault tolerance and availibity. Would highly recommend that you go have a look at Personalize and give it a go. See what I did there…