The Management of Direct Material Cost During New Product Development: A Case Study on the Application of Big Data, Machine Learning, and Target Costing

This dissertation thesis investigates the application of big data, machine learning, and the target costing approach for managing costs during new product development in the context of high product complexity and uncertainty. A longitudinal case study at a German car manufacturer is conducted to examine the topic. First, we conduct a systematic literature review, which analyzes use cases, issues, and benefits of big data and machine learning technology for the application in management accounting. Our review contributes to the literature by providing an overview about the specific aspects of both technologies that can be applied in managerial accounting. Further, we identify the specific issues and benefits of both technologies in the context management accounting. Second, we present a case study on the applicability of machine learning and big data technology for product cost estimation, focusing on the material costs of passenger cars. Our case study contributes to the literature by providing a novel approach
to increase the predictive accuracy of cost estimates of subsequent product generations, we show that the predictive accuracy is significantly larger when using big data sets, and we find that machine learning can outperform cost estimates from cost experts, or produce at least comparable results, even when dealing with highly complex products. ... mehrThird, we conduct an experimental study to investigate the trade-off between accuracy (predictive performance) and explainability (transparency and interpretability) of machine learning models in the context of product cost estimation. We empirically confirm the oftenimplied inverse relationship between both attributes from the perspective of cost experts.
Further, we show that the relative importance of explainability to accuracy perceived by cost experts is important when selecting between alternative machine learning models.
Then, we present four factors that significantly determine the perceived relative importance of explainability to accuracy. Fourth, we present a proprietary archival study to investigate the target costing approach in a complex product development context, which is characterized by product design interdependence and uncertainty about target cost difficulty. We find that target cost difficulty is related to more cost reduction performance during product development based on archival company data, and thereby complement results from earlier studies, which are based on experimental studies.
Further, we demonstrate that in a complex product development context, product design interdependence and uncertainty about target cost difficulty may both limit the effectiveness of target costing.