Performance pricing modeling process development by design science research
Härmä, Jiri (2023)
2023
Tietojohtamisen DI-ohjelma - Master's Programme in Information and Knowledge Management
Johtamisen ja talouden tiedekunta - Faculty of Management and Business
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2023-04-21
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202303293309
https://urn.fi/URN:NBN:fi:tuni-202303293309
Tiivistelmä
Multi-linear performance pricing models (MLPP) are widely used by organizations to analyze cost estimates for sourced parts in bulk. However, this analysis process is often resource-intensive, requiring significant tacit knowledge and resources from cost engineers. In response, this design science research (DSR) aims to increase the efficiency of this process by improving the coverage of MLPP models through the development of a research artifact.
In our research, we selected a design and development-centred entry point to identify the problem statement (PS). Through the development of a constrain-centric framework (CCF), we were able to break down the PS into a series of research questions (RQ) that could be addressed by implementing the research artifact.
To solve the PS, a machine learning (ML) solution using K-means clustering was utilized to identify parts similar to those parts used to model the existing MLPP models. The resulting similar dataset was then analysed by the existing MLPP model, resulting in a significant increase in coverage for the sheet metal MLPP model from 157 parts to 395 parts, representing a 151 % increase in the coverage. As a result, the sheet metal parts yearly annual spend analysed increased by over 73 %.
The research efforts not only resulted in an increased coverage of the MLPP model, but also in a significant improvement in the models mean absolute prediction accuracy (MAPE). Specifically, when the existing MLPP model was used to predict cost estimates on data gathered by the ML solution, we observed an improvement in the MAPE, in some cases halving the error in the prediction. The results suggest that the approach can be used to increase the coverage efficiently simultaneously improving the prediction accuracy.
Several important topics for future research related to MLPP models in cost management (CM) were identified by this research. These topics include the development of governance policies to ensure effective and manageable use of the MLPP models, as well as the implementation of life cycle management (LCM) strategies to optimize their usage over time. Additionally, there is a need to examine the scalability of these models across different commodities.
In our research, we selected a design and development-centred entry point to identify the problem statement (PS). Through the development of a constrain-centric framework (CCF), we were able to break down the PS into a series of research questions (RQ) that could be addressed by implementing the research artifact.
To solve the PS, a machine learning (ML) solution using K-means clustering was utilized to identify parts similar to those parts used to model the existing MLPP models. The resulting similar dataset was then analysed by the existing MLPP model, resulting in a significant increase in coverage for the sheet metal MLPP model from 157 parts to 395 parts, representing a 151 % increase in the coverage. As a result, the sheet metal parts yearly annual spend analysed increased by over 73 %.
The research efforts not only resulted in an increased coverage of the MLPP model, but also in a significant improvement in the models mean absolute prediction accuracy (MAPE). Specifically, when the existing MLPP model was used to predict cost estimates on data gathered by the ML solution, we observed an improvement in the MAPE, in some cases halving the error in the prediction. The results suggest that the approach can be used to increase the coverage efficiently simultaneously improving the prediction accuracy.
Several important topics for future research related to MLPP models in cost management (CM) were identified by this research. These topics include the development of governance policies to ensure effective and manageable use of the MLPP models, as well as the implementation of life cycle management (LCM) strategies to optimize their usage over time. Additionally, there is a need to examine the scalability of these models across different commodities.