Skip to main content
CUSTOMER STORY

Personalizing the beauty product shopping experience with data and AI

300

Millisecond predictions for real-time recommendations

200x

Faster time-to-market for new models

40%

Reduction in staff costs

flaconi hh header image color
CLOUD: AWS

"Thanks to Databricks, we have all the technology in place to scale and develop the best beauty experience for our customers in order to boost engagement and loyalty.”

— Dr. Carlotta Schatten, Data Science Team Lead, Flaconi

With over 55,000 products and more than 2.3 million customers, Flaconi wanted to leverage data and AI to become the No. 1 online beauty product destination in Europe. However, they struggled with massive volumes of streaming data and with infrastructure complexity that was resource-intensive and costly to scale. To tackle these issues, Flaconi implemented the Databricks Data Intelligence Platform to create a completely personalized online shopping experience for their customers. Through streaming analytics and machine learning, they are now able to provide real-time predictions to ensure rapid and accurate recommendations for customers — driving bigger cart values and, therefore, making a positive impact on conversion rates and revenues.

Scaling data analytics to align with business growth

Flaconi’s aim is to inspire their customers and guide them through a world of curated beauty brands and products. To get there, the company uses personalization and highly targeted recommendations based on customer behavior and historical purchasing data, allowing them to deliver an individually tailored, user-centric experience to every person who visits their site.

Dr. Carlotta Schatten, Data Science Team Lead at Flaconi, said, “Our main objective was to have a scalable infrastructure that would fulfill our vision and allow our teams to work together in an agile way to support the rapid growth of the company.”

The greatest challenges facing Flaconi were the increasing volumes of data, the difficulty of integrating existing tools and the inefficient refactoring of different parts of data products. The company needed a multifaceted solution: support for the ever-growing amount of data, improvements to the infrastructure and deployment strategy so that it worked synergistically, and increased flexibility to try out new technologies, such as deep learning, natural language processing, image processing and streaming analytics. They also needed to address the different infrastructure requirements of data analysts, whose focus is on reporting, and data scientists, who prefer to work with raw data. Flaconi selected the Databricks Data Intelligence Platform on AWS to meet these demands while also delivering a unified, scalable, collaborative lakehouse that would benefit all data teams.

A dedicated, scalable and flexible solution

“We decided to take one use case where we had a very strong vision, that was very user-centric and could make a significant difference for our customers,” Dr. Schatten said. “It was also perceived to be the most challenging in terms of data quantities. As we are increasing the number of data science use cases, we needed a solution that allows the teams to manage the different lifecycles efficiently while working synergistically with the data.”

Having introduced Databricks, Flaconi’s data teams are now able to enjoy smoother transitioning from development to production, making it easier and quicker to meet their code quality standards. Thanks to the scalability of the platform, the company is able to handle large volumes of streaming and batch data, while the platform’s versatility has allowed Flaconi to design an infrastructure that covers all their use cases.

Collaboration has also been improved, as Databricks makes it simple for teams to share any kind of code or issue via structured logging and versioning through interactive notebooks. Dr. Schatten commented, “Working together using lakehouse architecture is allowing us to better structure and optimize our ML-dedicated data extraction processes. In addition, we are able to integrate with Tableau to deliver business performance reports for our stakeholders. This improves transparency within the business.”

Specific components that have been applied at Flaconi include Delta Lake to build reliable streaming data pipelines and ML-dedicated data extraction processes, and MLflow for model lifecycle management, including versioning, model registry, easy-to-track experiments and results visualization. As an AWS shop, they are also using SageMaker and the DeepAR algorithm to forecast demand against historical trends.

Implementation has been fast and reliable, according to Dr. Schatten. “Databricks helped us to define a dedicated solution and provided in-depth workshops and sessions for our individual teams, combining all the different aspects into a working production-ready MVP. We achieved a working solution in a very short time as a result, putting the tools into context with many of our business use cases.”

Using Data + AI to drive customer engagement and revenue

The return on investment has been considerable for Flaconi from both a revenue and a cost-saving perspective. Databricks enabled infrastructure setup to be reduced from two days to 15 minutes without dependencies from other departments — allowing the ML team to deploy recommender models to customers 200x faster, which has improved customer engagement. The cost of the staff needed to develop their ML models has decreased by 40%, due to the reduction in engineering time required for developing infrastructure, optimizing code and automation. Predictions have been accelerated from 20 minutes to 300 milliseconds, resulting in significantly faster recommendations to customers, and therefore impacting conversion rates and the speed of purchasing decisions, therefore we expect increasing overall revenues with net order income increasing by 5%. The introduction of in-cart recommendations is also driving larger cart values, leading to increased customer lifetime value and share of wallet.

With 2.3 million active customers and data growing by 50% annually, Flaconi is now ready to scale their data operations as their business expands. Future use cases will include using advanced analytics techniques like natural language processing for various marketing applications and boosting customer lifetime value.

Dr. Schatten concluded, “Thanks to Databricks, we have all the technology in place to scale and develop the best beauty experience for our customers in order to boost engagement and loyalty.”

Databricks and Tableau

By unifying their data with the Databricks Data Intelligence Platform, Flaconi can monitor the performance of their machine learning models — delivered through performance reports via Tableau — for stakeholders to better understand how models are impacting the business.