Miroglio Group Gains Better Customer Insights Using Machine Learning on AWS with beSharp

Miroglio Group is a group of six Italian fashion brands with international operations. The company worked with AWS Partner beSharp to improve its customer loyalty program, make better use of personalized marketing and communication to increase customer loyalty, and reduce churn using machine learning solutions.

Fashion company Miroglio Group includes six brands with a strong and proud Italian identity (Motivi, Elena Mirò, Fiorella Rubino, Oltre, Luisa Viola, and Diana Gallesi). The company operates in several fields, including clothing brands, integrated logistics, and supply chain management.

Miroglio Group is committed to learning everything it can about the fashion industry to allow it to grow and innovate. To that end, the company wanted to make more use, and sense, of the data it was generating to help develop its brand loyalty scheme, improve how it communicated with its customers, and to reduce churn.

With structured and unstructured data flowing in from its complex operations—and only a small technical team—the company believed machine learning (ML) technology could help it meet its goals.

But early attempts to harness ML solutions were disappointing, producing results that could not be translated into action by the business. To help capture the potential of its data, Miroglio Group turned to AWS Partner beSharp to develop a solution on Amazon Web Services (AWS).

Miroglio Group needed a way to streamline and automate the process of creating machine learning algorithms to provide actionable results. beSharp focused initially on four of Miroglio Group’s six Italian brands.

The retailer wanted to divide customers into meaningful groups or clusters. This would then allow it to target marketing and other promotional activity at those groups with the intention of moving customers from less profitable to more profitable groups. This would represent a major step forward compared to its previous business intelligence reporting system.

To help clean and normalize data, beSharp first moved Miroglio Group’s customer data from a legacy on-premises database into a data lake built on Amazon Simple Storage Service (Amazon S3), a cloud-based, highly-scalable object storage service.

beSharp used AWS Glue to provide serverless data integration to discover, prepare, and combine data for analytics and machine learning. In addition, Miroglio Group uses Amazon Athena for detailed queries and Amazon QuickSight, a cloud-native, serverless business intelligence service for fast overview of store and product line performance.

With data ingestion, extraction, and storage resolved, beSharp and Miroglio Group turned their attention to creating ML models to segment customers into clusters based on spending habits, last purchase, and proximity to stores. The two companies chose Amazon SageMaker to build, train, and deploy the models.

Previously, segments were static and difficult to change, but using AWS the company can run a refresh once a month to see which customers have moved up or down the segments and begin to make sense of why. This more flexible approach is already showing success at increasing customer spending in higher level clusters. Although it is too early to talk about reducing churn rates or improving average individual customer spending, Miroglio Group now has the systems in place for that analysis and can develop new models faster.

The company can also be more granular with customer communications and design marketing campaigns with one cluster in mind, rather than just promoting a general message. “Our data lake means that we can now build both the predictive and prescriptive analytics applications that were previously impossible for us to create,” says Marcello Offi, business intelligence architecture and delivery manager at Miroglio Group. “We have reduced the time needed to develop analytics applications from months to just weeks.”

With a small data team Miroglio Group wanted to automate systems for ML model building. Amazon SageMaker makes building ML models more accessible with integrated development environments and no-code visual interfaces for non-technical business analysts. “The great thing about Amazon SageMaker is that you don’t need a strong data science background to build the kind of medium-strength models that we require,” says Offi. “It would be impossible for us to have achieved so much using any other data system that I’m aware of.”

Miroglio Group began working with beSharp following a suggestion from AWS. beSharp proved to be a true partner and collaborator and not just a supplier of services. “We didn’t want a typical partner—we wanted a partner to not just build but also to transfer knowledge,” says Offi. “beSharp did this perfectly—developing systems but also teaching us how they are built, how to make the most from them, and how to adjust them in the future.”

The transfer of knowledge meant the Miroglio Group team now has full confidence it its ability to run and improve the system in the future.

While Miroglio Group still has to make full use of its continuously improving ML system, it can already track customers more accurately across brands and different channels. The initial use of the channels has been to implement specific marketing campaigns designed to move customers to more valuable segments and has proved effective for customers already near the top of the 6 segments.

The next step is to roll out the system to the other two brands in the group. To this end, beSharp and Miroglio Group are building a set of applications to allow easier and more agile ingestion of data from both its own data lake and from Google Analytics.

Miroglio Group also plans to make both data and tools more widely available across the group. The team also has three specific business cases under consideration including one to create forecasts for ecommerce demand based on ML modelling.

Using AWS ML solutions, Miroglio Group is in a position to develop its loyalty scheme, improve communications, and reduce churn. “Today we can build, test, and implement highly targeted marketing strategies based on accurate customer clustering,” says Offi. “Thanks to beSharp and AWS we are working with the insights we need to serve our customers better.”