A choice behavior experiment with circular business models using machine learning and simulation modeling

Abstract

BackgroundTransitions from a linear (take-make-dispose) to a circular product system (considering reuse/remanufacturing/recycling) require changes of business models through new value propositions. Therefore, the focus for industrial businesses shifts from selling physical products to, for example, providing access to functionality through business innovation. In this context it is particularly challenging to understand what complexity a new concept like circular economy (CE) brings to established businesses where the success and the failure of the business is dependent on customer’s acceptance of new value propositions. ObjectiveThe objective of this paper is to develop an algorithm based on gathered survey data to “learn” choice behavior of a small customer group and then replicate that choice behavior on a larger population level. PurposeThis paper explores the opportunities of different circular business offers in the city of Stockholm by embedding support vector machine classifiers, which are trained on CE survey data, in a simulation model to quantify and study choice behavior on city level. MethodStated choices from CE surveys including unique demographic data from the respondents, i.e. age, income, gender and education, are used for algorithm training. Based on the survey data, support vector machine algorithms are trained to replicate the decision-making process of a small sample of respondents. The example of a washing machine is used as a case study with the attributes price and payment scheme, environmental friendliness as well as service level. The trained support vector machines are then implemented in a simulation model to simulate choice behavior on population level (Stockholm city). OriginalityThis paper is the first of its kind to use both machine learning and simulation approaches in a CE market acceptance context. FindingsBased on the washing machine-specific survey and Stockholm-specific customer data, results indicate that larger share of the Stockholm population would be willing to opt for circular washing machine offers compared to the existing linear sales model. Given the data-driven nature of machine learning algorithms and the process-oriented structure of simulations programs allows for generating large amounts of data from small samples. This supports exploration of new emerging areas like CE in addition to saving time and expenses.