This accelerator based on AWS Cloud Services is an accelerator for companies to develop comprehensive data strategies, extract value from their existing data and become a profoundly data-driven company. It is an MLOps based Well-Architected Framework (WAF). It accelerates the process of enhancing existing data within a company through the acquisition of additional data assets, allowing for more effective analysis of customers' shopping baskets and the creation of targeted, personalized offers.

The accelerator is dedicated to supporting a large variety of verticals and use cases and is easily extensible and customizable. It delivers its potential in existing customer relationships, where the customer is expected to get value from additional products from your portfolio, but there is no “proof”.

It has successfully been implemented with Hospitality Digital, a Metro company focusing on supporting the hospitality sector with competitive and innovative digital solutions. To learn more about this, we will soon publish a whitepaper (please reach out if you want to get early access).

Product overview

The concept behind the Hyper-Personalized Offerings Accelerator is to empower sales representatives with a data-driven solution that allows them to uncover previously unknown sales potential and thus to prepare hyper-personalized offers for each individual customer.

One example of such an end-customer is a restaurant owner. For this customer, the sales representative prepares individual offers. For the accelerator, we evaluated various types of data to create a comprehensive profile of the customer, including:

• Customer master data (provided by the vendor)
• Customer purchase history data (provided by the vendor)
• Menu data of the customer (data freely available on the internet, such as via Lieferando or restaurant homepage)

By utilizing the customer data, we can create a detailed profile of the customer, including information about their homepage, shopping history, and purchasing behavior. This data is then supplemented by customer-specific restaurant data, specifically menu items of the restaurant that are crawled from the freely accessible sources on the internet. Through machine learning, we match recipes to the menu items, ingredients to the recipes, and finally, articles offered by the customer to the ingredients.

Comparing the customer's desired ingredients with their purchase history allows us to identify gaps in their current purchase behavior. The gaps typically represent products sourced from a competitor.

To enable the sales representative to perform this gap analysis, a web application was developed to serve as a front end to the approach outlined above. The tool enhances the digital experience for the sales force and brings several functionalities:

• Information: The information page contains essential information about a customer including address details, ID, restaurant categories, digital tools and opening hours.
• Menu & Products: The menu and product categories represent the gap analysis and show all menu items of a restaurant with the corresponding ingredients and tag them with "recommended" or "buying".
• Benchmark Potential: The Benchmark Potential shows the different product groups of the customer and compares the average branch revenue with the customer's revenue. The goal is to identify possible gaps and show specific articles that can potentially be offered to the customer.
• Analytics: The Analytics function gives insights into the customer’s basket. Among other things, the average shopping cart value is displayed and the corresponding trend over the past months and all purchases can be viewed including the purchased products.

All in all, the accelerator is not only a sales support tool that serves as the front end for the Sales Representative, but also the holistic approach to create personalized offers for customers using various data assets.

Essential AWS Services that the accelerator is based upon:

• AWS S3 as Object Storage: Storing data, experiment meta-data and trained models.
• AWS Sagemaker Pipelines: ML training pipelines.
• AWS Sagemaker Endpoint: Inference endpoint for predictions.
• AWS Step functions and Lambda functions: Creates ETL pipelines to transform raw data for pre-processing.
• AWS Cloudwatch: Monitoring and log collection during training and applications.