What will my customers buy tomorrow? Taking the place of experience and gut feeling, IT solutions have long been a reliable predictor of future demand for goods based on a solid scientific foundation. One of the leading providers is software company Blue Yonder based in Karlsruhe, Germany.
How many strawberries will we sell on Friday? How many employees do I need in the next few days? For such predictions, most retail and logistics today still rely on a rather simple method: They primarily count on experience, belief and a little bit of gut instinct. But a change of thinking is gradually setting in. The magic word are big data. After all, data generated on a daily basis can indicate more than a positive correlation between summer temperatures and drink sales.
Blue Yonder provides an appropriate system for scientifically evaluating this data, and the company is taking a leading role within Germany in this context. Dr. Michael Feindt, who founded the company in 2008: “Even better predicting the next day or the next week can already bring substantial profit optimization for retailers and the entire supply chain.”
Scientifically analyzing complex systems
Before starting the Karlsruhe-based company, Feindt had worked in the European Organization for Nuclear Research, CERN, in Geneva. In 1997, he started working as a physics professor at the Karlsruhe Institute for Technology (KIT), where he is currently exempt from teaching obligations. According to Feindt, his primary scientific focus on elementary particles physics before founding Blue Yonder bears astonishing parallels to other complex systems such as can be found in business.
Michael Feindt: “As with weather forecasts, we can only make probability statements. But to stay with the metaphor: In many areas of business, ordering decisions continue to be made based on qualitative methods equivalent to Farmers’ Almanac weather forecasts and are not based on scientific methodology of modern weather forecasts. In this regard, even slightly better predictions can have a large, positive impact on the efficiency of many business processes.”
Innovations also must overcome resistance
Sales history provides the basis for the respective retail calculations (i.e., data that is readily available in every company). Based on a mathematical method, an algorithm takes this data and calculates which sales are expected under certain conditions. However, NeuroBayes, the algorithm applied by Blue Yonder, does not always follow the once-established pattern. The futuristic formula is actually capable of learning and automatically searches for corresponding correlations.
The program itself handles information about holidays or the weather. However, at the beginning, a lot of information is also manually entered, such as when a bus stop is moved, or a competitor starts a price promotion.
Integrating all aspects of the value chain is important in order to tap into the full potential of the mathematical prediction models. In this respect, Blue Yonder employees often work to convince their customers. For example, some buyers are afraid of having their freedom of choice limited. However, machine learning algorithms actually do offer scientific decision support, and the final decision (i.e., initiating the order process) can still rest with a buyer. But many retailers today are already starting to take it further, automating the entire process in order to make processes more efficient and profitable.
Example of returns management
Much of the optimization potential in logistics exists primarily in returns, which constitute an important cost factor for today’s online retailers. Apart from algorithmic fine-tuning, sometimes simple tools can already help. Michael Feindt: “The likelihood of return can often be significantly reduced by directly addressing the customer. For example, if two identical articles of clothing are ordered in different sizes and the customer is told that this product’s sizes are very standard, the customer usually decides not to order the second size. It is also helpful to appeal to the ecological sensibilities of the purchaser. Reasonable prices and realistic data about delivery times also reduce the rate of returns.”