The art of forecasting: Less intuition and more data for supply chain control

What company in the world doesn't want to predict the future? If it were possible, a company could predict what type and number of products it would sell (and thus how much it should buy), know how much a customer is willing to pay for each product, or figure out how to retain each user according to their preferences. Great business, whether it's for the corner fruit shop or for large multinational corporations. Fashion, whose products are largely perishable, has been dreaming up this idea for decades.

In the absence of a crystal ball, or other similar device, The art of forecasting has been a cornerstone for fashion companies to reduce cost margins Its operations can be economically viable. Although the reality of a knowable future is still far away, artificial intelligence (AI) is bringing companies increasingly closer to that goal, especially in everything that affects the supply chain.

According to McKinsey data. In the next three or five years AI can add between 150,000 million to 275,000 million to the fashion sector Euros of operating profit. This technology has allowed companies to create intelligent supply chains that, through predictive artificial intelligence and machine learning algorithms, collect and analyze large amounts of data on inventory and demand in real time and thus allow for more efficient decision-making.

Artificial Intelligence could add between $150,000 million and $275,000 million in operating profits to this sector.

The impact of these technologies on business profits may be so relevant that a decade ago companies in this sector invested only 1% of their sales in technological development, according to data from the technology company Nextail, which specializes in planning and inventory management for retailers. Today, this number has already risen to 2%, and investments are expected to double in the next five years to 4%..

In fact, by 2025, it is expected Up to 38% of companies worldwide have adopted AI in their supply chains as a “core” tool., according to data from OBS Business School, compared to the 11% who currently do so. “Fashion supply chains are clearly different from those of other sectors, and due to the type of product (much more seasonal and shorter lifespan) and other factors, they present greater complexity,” explains Analyticalways Founder and CEO, Amancio. junior.

The use of predictive AI in the sector is divided into two timelines. Pre-season or Before the start of the seasonwhen companies use technology to calculate factors such as expected demand For each product through data sets such as past purchasing trends. These calculations are made taking into account more than a hundred predictors (elements that determine the probability of demand), ranging from seasonal factors to response to similar elements.

Guaranteed, Predictive AI analyzes data sets from other years of these predictors and, through behavioral patterns that occurred in the past, attempts to predict future development.. In many cases, to increase the accuracy of the calculations, these algorithms must be “trained”, which is done through a process in which they input data from, say, 2022, to make the system make predictions for 2023. Once these calculations are made, the engineers By inputting actual data from the same year, the AI ​​compares its predictions to actual behavior, allowing it to learn.

These accounts have the potential to Avoid overcrowding, thus reducing unnecessary production costs Of items that will later be left in the warehouse. “In today’s business environment, companies face the need to effectively manage supply chain activities that increasingly extend beyond their borders,” the study conducted by Western Illinois University explains. Big data analytics in supply chain management. Predictive AI can help companies avoid situations such as bottlenecks, supply disruptions, or unexpected changes in demand.

“The main goal of predictive analytics is Increasing companies' efficiency and improving their financial results“, also confirms Daniel Martinez Pérez, Director of Desigual's Data Center of Excellence.

Other major uses of this technology include in the supply chain Presentation layout, so that it becomes more dynamic Avoiding the opposite situation, which is a lack of inventory, or improving the methods of sending goods, whether in terms of distance or other factors such as the condition and conditions of each route.

Once the season starts, The role of predictive AI is to “recalculate” all those predictions To adjust other factors such as the number of items in stores or the number of items Pricing. “If a product is sold in one store and not in another, and businesses are in a bad position, this has the effect where the product may end up being offered for sale, resulting in a loss of profit margin,” Junior explains.

“In the fashion sector, new products are constantly being launched, so There is a direct relationship with demand which requires almost weekly planning“Continues the founder of Analyticalways. For this, control mechanisms are used. Key Performance IndicatorsOr indicators performancethrough which the behavior of each item in different stores is continuously evaluated and calculates whether, for example, it would be more profitable to move this item to another store where it will achieve greater success.

The use of predictive AI in fashion is divided into pre-season moments and season moments

Pricing And customer segmentation

“Thanks to artificial intelligence We can predict, for example, the best price to apply to each product “Based on the inventory we have available, or the ideal discount path for each product,” Martinez further explains.

Dynamic pricing, or Real-time pricingthrough which the product can change its value according to changes in demand and supply, the total logistics cost or even competitors’ prices, and it is one of the latest applications of predictive artificial intelligence in this sector.

The predictive basis of these tools is what allows this Analyze how consumers respond to each price and calculate the “ideal price” based on different variables: Season of the year, origin of consumers, product success, etc. Along with the process known as “segmentation” or “profiling,” which allows for increased user customization, it indicates the type and quantity of products to be sent to each store.

“Data is very important for this sector, and has been collected for a long time. But in recent years, Our ability to predict trends has increased “Extrapolate this data to different areas of work,” Martinez explains.

But to implement this technology, A human team is needed to develop the necessary algorithms and models, which increases the demand for certain profiles Like Martinez's corporate work. As mentioned by the Data scientistYears have passed since full pages of Excel have welcomed the public landscape Machine learningPredictive artificial intelligence and big data.