The increasing prevalence of data collected on customer’s purchasing habits in recent years has led many companies to use this information for operational decisions, but decision-making can go awry without an appropriate analytical method to handle this insight.
Huanan Zhang, assistant professor of industrial engineering in the Harold and Inge Marcus Department of Industrial and Manufacturing Engineering at Penn State, received a $22,500 start-up grant from MHI, the nation’s largest material handling, logistics and supply chain association, to help improve operational decisions made based on gathered data.
Zhang’s research will focus on creating a data-driven algorithm for companies to raise their profits by improving their warehouse inventory forecasting, the way in which a company decides how they will stock the type and amount of their products. “Current operational decisions made based on historical data are like a circle: data, decision, implementation and effect,” Zhang said. “When managing a high-dimensional inventory system like a warehouse, implementing a misrepresented operational policy based on unchanged historical data can lead to a ‘spiral-down effect’ where both the quality of data being collected and operational decisions deteriorate over time. It’s a difficult domain to analyze due to the amount of data available and the amount of possible outcomes.”
By combining his expertise in operations research with his interests in data-driven algorithms and supply chain problems, Zhang will work to create an algorithm focusing on using customer transaction data to explore customer purchasing behavior. The algorithm will aim to automatically improve decisions to effectively balance learning and earning.
Zhang is particularly interested in urban warehouses, where a less-than-ideal algorithm becomes even more problematic due to the allotment of space. Urban warehouses, which are often smaller and more expensive than a traditional warehouse, are typically located within large metropolitan cities where space is sparse and strategic stocking is critical.
According to Zhang, urban warehouses rose in popularity due to same-day shipping from retailers. “If a company stocks an item in an urban warehouse and it gets a lot of sales, it might just be because of the location and fast shipping,” he said. “There isn’t a way to prove that the consumer enjoys or wants this specific item. We need better algorithms to explore product options for companies to make better selections in their urban warehouses because of the limited space.”
Zhang explained that a company may place an item in an urban warehouse because of its regional popularity. Likely, customers in the city will then purchase the item due to the convenience of fast shipping from the urban warehouse, rather than wait for a different item at a warehouse located farther away. Subsequently, this leads an algorithm to skew the data in favor of this item.
The company, viewing the data collected during this timeframe, will likely continue to stock the item in the urban warehouse based on consumer demands.
“Data is the new oil for the supply chain industry, but like oil, the unrefined data usually cannot be directly used,” Zhang said. “Correlation is not causation. This algorithm will be able to prove, under all conditions, that it is accurate and not fooled by data. This tactic of ‘learning smartly’ will help in reducing product waste and increasing company revenue.”