Deploy a mechanism to keep personalized recommendations up-to-date - Maintaining Personalized Experiences with Machine Learning

Deploy a mechanism to keep personalized recommendations up-to-date

Publication date: September 2021 (last update: January 2022)

The Maintaining Personalized Experiences with Machine Learning solution streamlines and accelerates development by providing automated pipeline construction; automated personalization model configuration, training, retraining, and deployment; improved visibility into model performance; and advanced error handling mechanisms.

This solution helps you build personalized experiences with Amazon Personalize for your product portfolio and provide real-time, curated experiences across digital channels. Implementing this solution aims to increase your user engagement metrics, clickthroughs, and conversion rates by providing up-to-date product recommendations, personalized product rerankings, user segmentation, and customized direct marketing.

Personalization opportunity exists in multiple areas along the customer journey including:

  • Discoverability - Helping consumers easily and quickly discover products and content

  • Acquisition and retention - Attracting and retaining consumers in a digital environment

  • Engagement - Understanding, measuring, and increasing time spent engaging with products and content

  • Efficiencies and revenue - Increasing average revenue per user

This AWS solution accelerates the development and deployment of personalization workloads by leveraging the functionalities of the Amazon Personalize service and streamlining productionization and maintenance of Amazon Personalize solutions. It provides end-to-end automation for Amazon Personalize through the entire lifecycle of a workload, which includes:

  • Automating the creation of Amazon Personalize recommenders, solutions, and solution versions to baseline model performance (by comparing user-configured model offline metrics over time).

  • Presenting the results in an Amazon CloudWatch dashboard.

  • Automating the scheduled retraining and update of Amazon Personalize solution versions to assess their performance over time, as typical personalization workloads benefit from full model retraining every one to five days.

  • Automating the scheduled creation of batch recommendations and batch user segmentation.

This implementation guide describes architectural considerations and configuration steps for deploying Maintaining Personalized Experiences with Machine Learning in the Amazon Web Services (AWS) Cloud. It includes links to an AWS CloudFormation template that launches and configures the AWS services required to deploy this AWS solution using AWS best practices for security and availability.

The guide is intended for IT architects, developers, DevOps, and data analysts who have practical experience architecting in the AWS Cloud.