Big data and its various applications open many new possibilities in manufacturing and supply chain management; no question about it. However, Big Data does not automatically lead to success. On the contrary, a “thoughtless” or uncontrolled use of data can lead to errors and incorrect decisions. Therefore, the distinction between correlation and causality is important.
In addition to intelligent algorithms that can calculate valuable information from big data, we primarily need professionals with expert knowledge to achieve truly valid results.
That this is of utmost importance can be observed in a series of supposed correlations, which Harvard student Tyler Vigen lists on his website. Using algorithms, he determines some curious scenarios. For example, his calculations show that the divorce rate in Maine, USA, correlates with the per capita consumption of margarine. His graph shows that if you want to save your marriage, you should eat less margarine. Another calculation shows that the consumption of mozzarella cheese influences the number of graduations of construction engineers – or vice versa. Even though some civil engineers may now want to cheer spontaneously and go eat some cheese, we can easily identify these examples as misleading and can draw an important conclusion: A correlation is not necessarily a causality. A correlation describes the pure connection of these two situations; a causality is the concrete relationship between cause and effect.
The aim is therefore to link data intelligently and turn it into reliable and logical information, which in turn can lead to success for companies. In the mozzarella case, we are all experts who can assess the credibility and relevance of this supposed connection. However, in economic contexts and planning situations in supply chain management, logistics, manufacturing and trade, the results of calculations and statistics are often more difficult to interpret – this requires specialized expertise.
Important correlations in supply chain planning
In companies, for example, the demand planning process in supply chain management is very complex. Finding out how many items of a product the customer will order tomorrow or next week, requires insight knowledge and – somehow – the ability to look into the future. Since this is impossible, forecasting methods can be used to help the planners. These forecasts are based on various factors: historical data, seasonal sales, product launches or new products, sales promotions, etc. Specialized algorithms can find patterns based on all of this data. With the help of these forecast calculations and the knowledge of experienced planners, a good outlook can be achieved at this point.
Markets are changing. The requirements for availability are growing and individualization and speed are increasingly becoming a competitive advantage. In order to be able to act cleverly in a changing world, you cannot plan “into the blue”. Your planning must be precise; otherwise, you work inefficiently and produce excess stocks or even have a stock out. Therefore, identifying the right causalities is important.
Identifying the right causalities with the help of algorithms and technical expertise
If we want to use big data and intelligent algorithms effectively in order to be able to reliably estimate the future demand for products, experts must first determine which influences are affecting sales (where the causalities really exist). These influencing factors can include obvious reasons such as sales campaigns, but also the weather or major current events. Just think of the upcoming soccer World Cup. If you leave this fact out of your planning, you may be giving away sales potential. The information about this causality between the World Cup and the increasing sales of beer, for example, must be provided to the intelligent algorithm. Then it can make very reliable predictions with the right data. And on this basis, the entire supply chain management process becomes more efficient and secure.
The human being as the expert
Big data alone will not get us anywhere. Today’s modern algorithms have been developed to a point that they can recognize a correlation. They are often already able to identify the correct causality from a data pool. However, Tyler Vigen’s examples show that there can also be mistakes. If the technical expert sets the right direction and keeps control, the machine can deliver better results than a human ever could. Therefore, sustainable supply chain management depends on a combination of big data, intelligent algorithms and human expertise.
