From conventional statistical forecasting methodologies to Machine Learning: guiding the transition through tools’ continuous improvement

The fashion and luxury industry is characterized by short product life cycles, high volatility and low predictability of product demand . Empirical evidences showed that the business growth can no longer rely on marketing and communication activities; although, a market-driven perspective is eagerly required in order to align all the departments and processes within the supply chain. An efficient forecasting system can be identified as one of the crucial elements to achieve the needed reactivity to keep up with the latest trends and be able to produce and deliver on time the right product, in the right shop. The forecasting system previously exploited by Louis Vuitton’s sales planner, Sakura, was no longer able to satisfy these requirements, therefore it was replaced by a more efficient forecasting system: iForecast. The implementation of this new tool allows to transmit a more accurate signal to production, which will consequently be more flexible towards customer demand. However, in dealing with such a VUCA environment, Louis Vuitton has adopted a far-sighted vision in order to increase forecast accuracy and agility even further, by testing the implementation of a Machine Learning forecasting model to improve iForecast performance. Machine Learning belongs to this new generation software applications which are supporting real-time omnichannel processes allowing a global reach connectivity to multiple sources of data. Therefore, it allows to convert events within the product-customer environment into variables which will consequently impact forecast. Nevertheless, if on one side it leads to countless improvement, on the other side it’s relevant to notice that the transition from a tool to a new one always arises bias and rigidities among key users, sales planners in this case. Therefore, two main dashboards have been developed along this project in order to both outweigh the lack of a core reporting tool and allowing to have a transversal vision of KPIs among all the departments involved within the process.