How a grocery store offers a personalized experience with recipes | Amazon Web Services Blog

The grocery market was severely disrupted by the COVID pandemic. Consumers have shifted to online shopping and home delivery. For example, in 2020, the UK's share of total food products in the UK exceeded its pre-COVID forecasts in 2024, reaching 8.8% in 2020. Similar growth has been seen in the international market. The restrictions are gradually being lifted, but some consumer trends remain here. According to FoodNavigator, 91% of people in the UK plan to cook "as much or more" this year, and McKinsey reports that consumers are more likely to eat post-pandemic food than before. Expected to spend money.

Differentiating and reinventing the customer journey is critical to increasing the growth and efficiency of retailers, given the consumer's nesting and continued home cooking, as well as the small margins of the grocery store. Grocery stores have a competitive advantage because they collect customer data over time, such as purchase history, customer profiles and preferences, and loyalty-based information. Using this data to design and deliver highly personalized services not only improves customer satisfaction and loyalty, but also increases total sales by 1-2%. AWS's perspective on how retailers can use personalization to increase sales by gaining a wealth of insights is explained in the "Personalization: To Gain Deeper Insights and Increase Sales" post.

This article will show you how to provide a customized customer journey for online shopping. We believe it is important to personalize the customer experience in all retail categories, using solutions designed for grocery retailers as an example. Now consider a customer who buys groceries online. When you add an item to your shopping cart, you'll see a list of recipes that contain the same ingredients as the added item. The suggested recipes are personalized to the customer. They are based on customer profiles and past product preferences. From the consumer's point of view, personalized suggestions provide dietary inspiration and help with dietary planning. Meal kits are also popular with consumers. From the above, we will improve customer satisfaction and loyalty by proposing an essentially personalized meal kit that is seamlessly integrated with the flow of online grocery shopping. For grocery stores, apart from the benefits to the customer experience, consumers can add missing ingredients to their recipes based on recommendations, resulting in increased sales.

About the challenge of providing a personalized experience

When we discuss personalization with his AWS customers, they list some common barriers to recruitment. It integrates customer data from numerous data silos, modern data platforms, analytical frameworks and data science employee skills, and commercialization of personalized products. The difficulty of managing and maintaining the above solutions yourself can be solved by using AWS Managed Services (AMS). AMS helps you operate your AWS infrastructure efficiently and securely. In addition, developers can build services to provide customized recommendations without prior knowledge of data science. It's important for retailers that the operational complexity of managing the infrastructure does not contribute to revenue. Engineering teams can focus on their products and customers with solutions that consist of managed or serverless services.

In addition, the solution is designed to provide personalized recipe recommendations as customers update their shopping cart and integrate with retailers' online shops to enhance their customer journey.

At the heart of the solution is Amazon Personalize. Amazon Personalize is ideal for retailers who do not necessarily have in-house machine learning (ML) expertise. This allows developers to build applications with the same ML technology that Amazon.com uses for real-time personalized recommendations. In addition, Amazon Personalize is a managed service, eliminating the need for a team of retailers to build, maintain, or manage the infrastructure. It's all managed and abstracted by AWS.

The solution simulated two microservice patterns and built two services on AWS Fargate: a cart and a recommendation service. AWS Fargate is a serverless, pay-as-you-go computing engine. As with Amazon Personalize, using serverless services means that engineering teams can work on higher-value jobs rather than managing the server.

For more information on architectural design, see:

front end

Build artifacts are stored in Amazon Simple Storage Service (Amazon S3) buckets that hold your web application assets (product / recipe photos, web graphics, and so on). Amazon CloudFront caches front-end content from Amazon S3 and presents your application to users through the CloudFront distribution. The front end interacts with Amazon Cognito, which is used for all authentication requests.

Backend

The core of the back-end infrastructure consists of Amazon Elastic Container Service (Amazon ECS) and his AWS Fargate-hosted microservices. These represent cart domains and recommendation services. A cartUpdatedEvent is sent over his EventBridge bus every time a consumer updates their shopping cart, and this event is stored on Amazon S3 using Amazon Kinesis Data Firehouse. Once the event enters Amazon S3, it can be used to train the recommendation engine model and ultimately build historical datasets for future use cases. The recommendation engine leverages Amazon Personalize, a fully managed ML service that trains, tunes, and deploys custom ML models to provide customers with highly customized recommendations. In this solution, the web app queries the recipe service with various product combination samples in the shopping cart and returns a list of recipes that satisfy it. This list of recommended recipes is ranked by the degree of matching with the customer's past purchase-based preferences. Then a list is displayed and each customer is uniquely ordered.

Recipes The AWS Lambda function retrieves a recipe stored in Amazon OpenSearch Service (the successor to Amazon Elasticsearch Service) and displays it on the front end.

Below is a screenshot showing how the customer journey works when this solution is implemented in a grocery store web store.

Step 1

Consumers accessing the online shop visit the home page of the application (see below). Since this is the first entry point, there is no input to the recommendation engine. Therefore, the backend only displays random recipes.

Step 2

For example, consumers have decided to add pasta to their shopping cart.

Step 3

Recipes displayed to consumers on the landing page have been adjusted to only recipes that include pasta.

Step 4

Let's add more tomatoes and eggs to the shopping cart.

Step 5

When you return to the landing page, the recipe you see has changed to a recipe that also includes the two items added to your shopping cart.

Step 6

When the consumer moves to the shopping cart, the recipe is also displayed in the left panel. These recommended recipes follow consumers throughout the shopping process.

The backend takes three different ingredients, pasta, eggs, and tomatoes as inputs, builds various combinations of them, and ranks them based on their match with the appropriate recipe.

This solution lowers the barriers to entry in implementing a personalized experience throughout the customer journey. Amazon Personalize is used by retailers to enhance innovative use cases. You can see customer feedback and reference materials here.

If you're interested in discussing details, experiments, specific use cases and how to adapt to your customers with this solution, contact your AWS account team today.

The translation was done by Solutions Architect Saito. The original text is here.

About the author of this blog

Chara Gravani

Chara Gravani is a Senior Solutions Architect at AWS, supporting and driving cloud migration for enterprise retail customers in the UK. She specializes in data and analytics, and is passionate about D & I and building STEAM education accessible to all children.

Stefano Vozza

Stefano Vozza is an AWS solution engineer based in the United Kingdom. Stefano works with AWS Solutions Architects to create standardized tools, sample code, demonstrations, and quickstarts. He is also the backend developer for AWS Perspective.