Industry: Fashion | Year: 2018 – ongoing
Spindox supported a well-known Italian clothing company for men, women and children, with around 1500 stores in Italy and abroad, in the reorganization of their logistics and supply chain processes in support of both the e-commerce boom during the post-COVID-19 phase as well as in the optimization in facing fashion seasonality.
The crisis caused by the 2020 pandemic produced two different effects in the clothing retail industry. On one hand, it accelerated the transformations already underway, leading, in particular, to the boom of e-commerce, even in a country like Italy, which is notoriously behind. On the other hand, it has become more difficult to predict the changes in demand and therefore manage the phenomenon of apparel seasonality. Ublique is a suite of decision-making solutions that implement statistical tools and applied artificial intelligence to simulate, predict, and optimize processes.
To meet the needs of the client, we customized some of the modules in the suite and turned them into the essential elements of a profound transformation of the processes. There are two objectives, in particular, guiding the initiative: standardize the seasonal distribution and intelligently perform replenishment. This project is directly linked to another initiative underway for the same client, aimed at increasing efficiency and speed in the logistics and delivery processes of merchandise.
The solutions implemented and described below are scalable, can be used in a cloud infrastructure, and can create prediction models and simulation scenarios. The user always has the ability to confirm the implementation of an operational decision.
Challenge 1: optimization and uniformity of seasonal distribution to increase sell-through
Ublique | Demand Intelligence
Demand Intelligence is a solution composed of several functional modules. At the crux of this solution is the optimizer, which optimizes based on sales forecasts and store budgets and supports the Supply Chain departments, providing input on how to distribute the merchandise to the various stores. The module consists of an algorithm that considers the qualitative and quantitative constraints that include the customer target, type and sales budget of the store, minimum shipping parameters to minimize transport costs, sales target of each product, and stock availability.
- increase the sales capacity of the marginal stores;
- increase the sell-through, i.e., the ratio between the merchandise sold and the merchandise placed in the store. In practice, the sales percentage is maximized to reduce the stock inventory to a minimum;
- quickly and systematically evaluate the distribution of merchandise. This means that the distribution does not depend on a personal choice of an operator but on global corporate logic for all stores;
- reduce the burden associated with handling activities.
Challenge 2: “there is only one stock” – logical unification of the warehouses, optimization of the distribution performance and replenishment of the stores
The forecasting models implemented to address Challenge 1 require periodic review over the months. To better understand this, let’s consider a typical store. The usual procedure requires the store to order the merchandise for the following season about eight months in advance. During this time, sales performance may vary. This means that the order may not necessarily correspond to the actual demand of the consumers and may not be suitable for situations that can suddenly arise. Take for example the COVID-19 scenario. When the stores reopened after the lockdown, they found themselves with a large amount of merchandise in stock. At the same time, the sales forecasts that had been made at the beginning of the season had completely changed and no one knew how sales would go. This is where Spindox comes in. Spindox corrects these sales forecasts. The fact remains that the change in the scenario with respect to the initial forecasts can generate the large risk of decreased sales. It is, therefore, necessary to have a system capable of ensuring a real-time update of sales data to achieve flexible management of the merchandise in the warehouse and arrange for deliveries that meet the actual needs for the sales of each store. To meet this need, Spindox has developed a system that, rather than assigning the merchandise received in the warehouse in advance to the various stores, it records the entire inventory and automatically sorts it among the various stores according to an intelligent criterion.
Ublique | Forecast module of the Demand Intelligence solution
The Demand Intelligence forecasting module was used to provide the sales forecasts of each store and of each item/size for the near future (within 6/8 weeks maximum) through data that includes the personal data and sales information of the individual stores for the last four years. Specifically, the Ublique algorithm analyzes the sales data of the stores and provides a forecast of which products will be sold the most at a particular store, indicating the type and size of the product. In this way, the system is able to optimize the merchandise received by the warehouse by providing information on the type and quantity of merchandise to be delivered and to which point of sale.
Through these predictive models, it is possible to manage stock-outs by recalibrating the sales forecasts at the beginning of the season, addressed through Challenge 1 mentioned previously. These forecast models become the input data for the second step of the replenishment, which we have optimized in the following way.
The solution elaborates the scenarios for the redistribution of merchandise according to the new forecasts. This system not only determines the shipment and the type of merchandise to be transferred from the main warehouse to the other stores, but also provides information on the best way to transfer the merchandise from one store to another to increase the probability of sales. The scenario for the merchandise transfer from one store to another is only suggested after a global consideration of the costs. For example, if a small number of items is to be shipped from one store to another to increase the probability of sales but the costs for shipping and handling are too high compared to the profit that can be obtained from the eventual sale, the operation is considered not advantageous and therefore the scenario will not be proposed.
- Speed: the forecasting and optimization processes are completed within a few hours, despite the large amount of data used (the system analyzes about 60 million items);
- reduction of lost sales that are a result of stock inefficiency or stock-out;
- optimization of store replenishment considering efficient quantities of production batches, order tolerances, and existing stock;
- possibility to modify a decision made and to vary orders based on real-time information regarding the demand and availability of merchandise.