VistaPrint, a Cimpress company, is a design and marketing partner to millions of small businesses globally, offering marketing products such as promotional materials, signage, and print advertising. Over its more than 20 years of operation, VistaPrint has developed a cloud-native system to better comprehend its customers’ needs and offer personalized product recommendations.
Earlier, VistaPrint had an in-house developed product recommendation system, but it could not automatically scale up to meet increased demand. Modifying this system was time-consuming due to the required high level of machine learning and e-commerce expertise. These challenges led VistaPrint to design a new, scalable cloud-native system consisting of serverless and software as a service (SaaS) components. This system externalizes much domain-specific functionality, enabling easier operations and faster market changes.
To provide personalized product recommendations, VistaPrint first collects historical data in a warehouse from upstream platforms such as customer data platforms, order management, product catalog, and user management systems. After transforming the data to create training data for Amazon Personalize, VistaPrint imports this bulk historical data to train machine learning models. The company uses different methods to predict items that a user will interact with based on previous interactions and to generate recommendations for similar items. In order to maintain model relevance, the data updating and training processes are repeated regularly.
The company also streams ecommerce website events to a customer data platform (CDP) to capture website interactions, like viewing or adding a product to the shopping cart. The CDP collaborates with Amazon Personalize to generate real-time product recommendations as customers navigate the website.
Additionally, VistaPrint has developed a placement and offer engine, enabling data scientists and marketers to collaborate. This engine allows marketers to select a model, the visual style, and extra features. The same system is used to design and place product recommendations in email marketing campaigns.
This new system has increased VistaPrint’s conversion rate by 10% and decreased their total cost of ownership by 30%, compared to the old in-house solution. Amazon Personalize, a machine learning powered tool, is at the center of this personalized recommendation system. By aggregating data from all customer touchpoints across various business domains, a CDP like Twilio Segment allows VistaPrint and similar companies to create a cohesive view of their customers, facilitating more accurate and personalized product recommendations. VistaPrint now plans to read more on their data mesh strategy and more about cross-channel customer experiences with Amazon SageMaker, Amazon Personalize, and Twilio Segment. The authors of this article are Ethan Fahy, Mouloud Lounaci, Emeline Escolivet, and Vibhusheet Tripathi.
