What models are best for dealing with complex scenarios?
Companies and government organizations have access to growing volumes of data on which to base their decision-making. Sandjai Bhulai, professor of Business Analytics at VU Amsterdam, develops models that can help these organizations to process the information contained in these growing databases to allow them to achieve their objectives, such as reducing food wastage by means of smart logistics or saving water with the aid of dynamic pricing schemes.
Loyalty cards provide supermarkets with masses of new information minute by minute, on such topics as daily sales of bread, vegetables and fruit, the age of the purchaser and even the time of purchase. “This data allows supermarkets to refine their logistic planning,” Sandjai Bhulai explains, “to ensure that there are always just enough stocks of all products on the shelves, not too little and not too much.”
Sandjai Bhulai has other examples of the uses to which such data can be put. “Thanks to the presence of growing data streams on the weather and consumer preferences, water companies can make the price of water they supply dependent on its availability. This ensures for example that people use less water in dry periods, because it is more expensive then. Another example: social media such as Facebook and LinkedIn enable personnel officers to fine-tune the pay and conditions offered in their job adverts to the availability of specific types of workers.”
Prof. Bhulai’s research group builds three types of models to support organizations in their decision-making. First, the researchers use statistical methods to analyse the data and reveal significant correlations, for example between the time people buy bread and the age of the consumers involved. Secondly, they calculate the relevant costs, for example of food wastage in future resulting from the company’s current marketing policy. And finally, they compare the results of different scenarios through optimization.”
Tens of thousands of variables
“We use tens of thousands of variables in our calculations,” says Prof. Bhulai. “Supermarkets may have 50,000 or more different products, which they can order in different combinations or decide not to order. Completely new algorithms are required to calculate the consequences of these marketing policies. We also have to process the data to remove irrelevant associations as far as possible. And we often have to simplify the problem – and, of course, to calculate the effect of this simplification on the results.”
Planning and logistics
Prof. Bhulai’s research group has developed planning and logistics models for hospitals, police and fire services. In 2012, he and a number of colleagues set up the Amsterdam Centre for Business Analytics (ACBA), which has already made a name for itself. In this interdisciplinary centre, which covers the fields of mathematics, IT and business administration, research assistants studying for Master’s degrees and doctorates work on a wide range of problems for all sorts of companies. “The ACBA is a meeting place for research and business,” Prof. Bhulai says. “Our researchers take data provided by companies and use it to build models that help the companies to make the necessary decisions. And in the process, they earn themselves a Master’s degree or a PhD.”