The world of fashion is experimenting with artificial intelligence techniques for the retail sector and e-commerce sales, but there is no application regarding the sector chain.
We are convinced that, through network data, it is possible to know the tastes and behavior of the customer, in the same way through the transfer of company data, defined, for each product class, which is the preferential production channel and the best combination between supplier of raw material and supplier of internal or external processing.
Our project applies artificial intelligence techniques in the specific sector of the fashion supply chain.
By analyzing the largest possible volume of data, we look for the correlations that can emerge between apparently uneven data such as production volumes, variety of models, production areas, suppliers, product complexity, raw material, quality control results.
Digital transformation is a topic that has now started and today companies have understood the importance of using the result of digitization, the data, in processes that find an "intelligent" interpretation, which is then useful to suggest the correct transformation of processes and behaviors. .
Artificial intelligence represents an additional tool in this race for efficiency which is raising the bar of competitiveness on more and more markets. In fashion, the most important luxury brands experiment with new technologies in the retail sector and use AI to learn more about the behavior and tastes of their customers, in stores and online.
However, these experiments have not yet been extended to the supply chain of the sector, so we decided to start with the first exploratory projects. We are convinced that through the analysis of company data we can define, for each product class, which is the preferential production channel and the best combination between raw material supplier and internal or external processing supplier.
With the information available, deriving mainly from the ERP, we look for the correlations that can emerge between apparently uneven data such as production volumes, variety of models, production areas, suppliers, product complexity, raw material, quality control results. To do this, we have therefore designed an architecture that conveys all this information within a "container" that orders and formats the data that will be analyzed.
The project is divided into five phases.
1 Design "Stealth Staging Area" identification of volumes and pre-selection of the original data present in Stealth
2 Architecture: Stealth Staging Area, Communication flows, Oracle Autonomus Cloud
3 Data mining: Analysis of extracted data and implicit information to make them usable and correctly formatted.
4 Machine learing: Exploration of the formatted data, in order to identify random patterns, trends and correlations currently unknown
5 AI: Application of AI techniques that use correlations, to independently make productive choices.
Our in-depth theme will be dedicated to the first two phases of the "Stealth Staging Area" and architecture design.
The first necessary phase was the selection of data from the great availability that the Stealth platform offers. At the same time, we had to perform structural transformations of a data that arises according to different logics than the final use of a machine learning system.
Among all the information that Stealth makes available to finalize a complex production, we report the main ones that make up the database of our project: bill of materials of the product, number of components, classes of use, processing phases, processing suppliers, classification of raw materials, production areas, product complexity.
The complexity of the product has led us to leave out the characteristics definable by human sensitivity and to devote ourselves to those deductible from information provided by the system, establishing a ranking of complexity to be included in the data set.
Our project is taking place in these months, and we are working on the data mining phase, using Oracle Autonomous Cloud interfaced with Stealth. We are available to welcome other companies that want to participate in this experimentation phase ensuring the confidentiality of information. With more data to analyze, you will have more indications and suggestions to improve the learning phase and achieve more targeted results.
Santo Lombardo, holds the role of Consultant Executive Fashion & Retail for Dedagroup Stealth. He developed and consolidated his professional experience working for the most important luxury brands. It accompanies fashion companies in their delicate phase of business transformation and implementation of the Stealth ® platform