Amazon Personalize is a low-code machine learning (ML) service that can generate custom recommendations through an application program interface (API) call for any application running on Amazon Web Services (AWS) infrastructure. The goal of Amazon Personalize is to deliver customized recommendations that will improve customer engagement.
Developers typically use Personalize to tailor product recommendations, content recommendations, search results and marketing promotions. Personalize is popular with developers who work on e-commerce sites because it allows development teams without technical ML experience to customize results for the apps they create.
The developer is responsible for providing training data, but Amazon is responsible for selecting the right algorithm, training and updating the AI model, and correlating the accuracy of the metrics. According to Amazon, this approach reduces the time it takes to build a machine learning model for recommendations from months to days. The service can use historical data stored in Amazon S3 as well as streaming data from apps to tailor results.
Pricing for Amazon Personalize is based on the size of training data, the amount of training time and the number of recommendations generated per hour.
How Amazon Personalize works
Amazon Personalize is based on the same technology Amazon Web Services (AWS) has been using for over twenty years and is accessed with an AWS console.The following steps must be carried out in order to implement personalized customer recommendations:
- Data must be formatted and input into the service. Inventory and user demographic information can be taken from an Amazon S3 bucket or an Amazon Personalize API can be set up to stream event or activity data, such as clicks, page views and purchases.
- Recommendation data should also be provided to the service. This includes any contextual information that might be relevant and an inventory of items that can be recommended, from articles to products to media.
- Amazon Personalize processes and examines the data in order to identify what is important. An algorithm is then chosen to train and optimize a personalization solution that is tailored to an organization’s data.
- The solution, or trained model, is deployed and implemented into applications through an API call. Potential integrations include websites, mobile apps, social media platforms, content management systems (CMS) and email marketing software.
Applications for Amazon Personalize
Relevant recommendations can be implemented in real time in a variety of use cases, including:
- Personalized recommendations - Targeted suggestions can range from next steps to product recommendations.
- Custom search - In order to improve customer experience (CX), Amazon Personalize can be used to design search functions that rank results based on each user’s behavioral data and preferences.
- Relevant notifications - Customers that only receive the most relevant marketing materials are more likely to convert. Amazon Personalize can ensure only appropriate, adapted notifications reach users.