Les Galeries Lafayette: Innovation and AI Serving Customer Satisfaction
As the leading French department store chain, Les Galeries Lafayette chose innovation to better understand the challenges of customer satisfaction.
Les Galeries Lafayette turned to Onepoint and its expertise in consulting and innovation to apply new methods of statistical analysis and artificial intelligence to assess customer satisfaction and shopping journeys.
This new approach aims to reconcile statistics and causality, offering a fresh perspective on customer experience analysis.
Our goal was to help our teams understand the value of enhancing customer satisfaction. The objective was to quantitatively support existing observations related to customer satisfaction. We wanted to use data to ensure that these observations were not just another opinion, but a solid, evidence-based perspective.
Les Galeries Lafayette chose to collect customer feedback through responses to a series of post-purchase questionnaires. As a buyer’s journey is influenced by multiple factors, the group needed a solution to efficiently analyze and leverage all the available data.
Thus, Les Galeries Lafayette turned to Onepoint to provide key performance indicators that would prioritize the customer experience roadmap. The challenge here was that the analysis results had to quantitatively justify the resources allocated to improving the customer experience.
Several international and cross-industry statistical studies highlight the connection between customer experience and business performance. These findings were theorized in the 1990s through the concept of the “experience economy.” The challenge of our support was to identify the key elements of this theory and apply them on a larger scale within Les Galeries Lafayette.
Complexity and interconnection of the issues to explore
Until now, our actions were prioritized based on the volume of customer feedback, without considering the potential value it could generate. We wanted to change our perspective by viewing the customer as the primary decision-maker of value, which led us to explore the connections between their feedback and their purchasing behavior.
To best target its areas for improvement, Les Galeries Lafayette needed to study the impact of different elements of the customer journey on their purchasing behavior. How can they best enhance the customer experience? What factors influenced the average basket size, and to what extent? How can the interconnected impacts of this list of factors be untangled?
Moreover, analyzing the responses posed an additional challenge. Saying that correlation is not causation is one thing. Being able to provide results that lean more towards the latter than the former is another.
Research at Onepoint, a Source of Innovation
The Institute, which supports Research and Entrepreneurial Innovation at Onepoint, offers a set of tools derived from research that help closely meet client needs.
Far from working in isolation, the methods used by our researchers allow for the integration of business expert knowledge throughout the study. This wasn’t about blindly trusting a model processing a set of data; rather, it was about co-creating a solution that would be both explainable and explanatory.
A Statistical Model Focused on Causality
When analyzing customer expectations, it’s important to consider the context of those expectations. A customer looking to buy lipstick would not have the same expectations for their experience at Les Galeries Lafayette as a customer looking to buy luxury clothing, for example. It’s crucial to consider all the data related to a problem to understand the dynamics that might connect them, especially when dealing with a problem where the multiplicity of information lies in customer profiles, the number of stores involved, and the vast range of products offered.
This is where the business expert comes in, using their knowledge to refine the model on which the analysis is based. The involvement of the business expert is one of the strengths of causal inference, a method halfway between statistics and machine learning within artificial intelligence.
This method stands out from traditional machine learning approaches precisely because it requires the expert’s input to bring their worldview into the analysis. This structure enabled it to establish itself as a reference for moving from simple correlations to causal links.
The use of external knowledge is reflected here through the definition of a knowledge graph, which was created by Les Galeries Lafayette in collaboration with Onepoint. This graph best reflects the French department store leader’s knowledge of its customer base. Once established, the graph serves as a guide for a statistical model that can then study all the available data.
Creating an interaction between my business expertise and their data methodology allowed us to move forward through successive iterations, enabling us to learn as we went. For example, this series of iterations allowed us to discard certain hypotheses that seemed promising but ultimately turned out to be dead ends. In this sense, I found this method to be more agile than traditional methods, which operate on a single cycle of request and result.
And After
Once trained, the analysis model provides a detailed description of the impact of a set of factors on a target value, in this case, the average customer basket. Each factor can then be analyzed individually based on its specific impact.
The results of this project allowed me to engage our teams on the topics highlighted by the analysis. Quantifying elements of the customer experience, such as the salesperson’s greeting, gives us access to a set of indicators. These indicators will be used to refine the prioritization of our actions to improve the customer experience.
Conducting this analysis allowed for the identification of key areas for improvement in terms of user experience, while quantifying their impact on the average customer basket. In doing so, Les Galeries Lafayette gained detailed results that will enable them to prioritize their future initiatives.