Designing a business model for machine learning based predictive maintenance

Abstract

This Master’s Thesis examined the predictive maintenance industry and focused on designing a business model for machine learning based predictive maintenance. A machine learning-based start-up company offered intelligent energy analytics for predictive maintenance and wants to enter the pulp and paper industry. However, the organization had limited knowledge in how they could plan for identifying potential sources of revenue, the intended customer base and how their product compared to existing products. Their business areas were energy efficiency and predictive maintenance. The purpose was to design a suitable business model that included these aspects and to position the proposal against current state-of-theart practice in the predictive maintenance industry. The study analyzed existing predictive maintenance solutions, the current business model and how it performs in comparison to competitors, and potential opportunities and threats. This was to acquire as much information as possible prior to the business model concept generation and screening stages. Data was collected from interviews and a literature review. Several analysis were carried out to investigate external market factors and internal company factors, to identify essential constituents that had to be considered during the concept generation phase. By using the analyses as foundation, five different business model concepts were developed for the industrial partner. During the screening phase, the business model concepts were evaluated in comparison to the company’s existing business model. Several models, hypothetically, outperformed the current business model. However, after elaborate reflections of these business models, it was necessary to terminate some of the concepts due to challenges related to a realistic implementation. The motive was that the eliminated concepts would have required a major resource allocation and since the company is a start-up, it has access to limited resources, thus restricting the available options. Consequently, only two business model concepts were selected as realistic suggestions. These models were discussed with the company for validation, which resulted in one of them being terminated but also in an additional business concept being generated. The new concept was combined with the remaining one, which became the final business model recommendation. The study resulted in a business model concept that was derived from the conducted analyses and defined criteria. The concept focuses on adding value to the machine learning based start-up and its customers through sensors for extracting the data and licensing their software to an external party’s platform to facilitate operations and reduce system complexity. As the concept had already been validated by the organization, a technology roadmap was established to provide detailed information regarding how the company could implement the business model suggestion in practice. Since the initial objective was to design a suitable business model and identify revenue sources, customer segment and existing products, the study achieved its original purpose. The Thesis finished with a discussion on the elicited business model, positioning it against competing businesses that applied state-of-the-art practices for maintenance management. The report then ends with a conclusion and the authors’ input regarding future research.