Strategic Project Portfolio Management by Predicting Project Performance and Estimating Strategic Fit

Loading...
Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

School of Business | Master's thesis

Authors

Åström, Joona

Date

2020

Major/Subject

Mcode

Degree programme

Information and Service Management (ISM)

Language

en

Pages

77 + 18

Series

Abstract

Candidate project selections are extremely crucial for infrastructure construction companies. First, they determine how well the planned strategy will be realized during the following years. If the selected projects do not align with the competences of the organization major losses can occur during the projects’ execution phase. Second, participating in tendering competitions is costly manual labour and losing the bid directly increase the overhead costs of the organization. Still, contractors rarely utilize statistical methods to select projects that are more likely to be successful. In response to these two issues, a tool for project portfolio selection phase was developed based on existing literature about strategic fit estimation and project performance prediction. One way to define the strategic fit of a project is to evaluate the alignment between the characteristics of a project to the strategic objectives of an organisation. Project performance on the other-hand can be measured with various financial, technical, production, risk or human-resource related criteria. Depending on which measure is highlighted, the likelihood of succeeding with regards to a performance measure can be predicted with numerous machine learning methods of which decision trees were used in this study. By combining the strategic fit and likelihood of success measures, a two-by-two matrix was formed. The matrix can be used to categorize the project opportunities into four categories, ignore, analyse, cash-in and focus, that can guide candidate project selections. To test and demonstrate the performance of the matrix, the case company’s CRM data was used to estimate strategic fit and likelihood of succeeding in tendering competitions. First, the projects were plotted on the matrix and their position and accuracy was analysed per quartile. Afterwards, the project selections were simulated and compared against the case company’s real selections during a six-month period. The first implication after plotting the projects on the matrix was that only a handful of projects were positioned in the focus category of the matrix, which indicates a discrepancy between the planned strategy and the competences of the case company in tendering competitions. Second, the tendering competition outcomes were easier to predict in the low strategic fit quartiles as the project selections in them were more accurate than in the high strategic fit categories. Finally, the matrix also quite accurately filtered the worst low strategic fit projects out from the market. The simulation was done in two stages. First, by emphasizing the likelihood of success predictions the matrix increased the hit rate and average strategic fit of the selected project portfolio. When strategic fit values were emphasized on the other hand, the simulation did not yield useful results. The study contributes to the project portfolio management literature by developing a practice-oriented tool that emphasizes the strategical and statistical perspectives of the candidate project selection phase.

Description

Thesis advisor

Liesiö, Juuso
Vilkkumaa, Eeva

Keywords

project portfolio management, strategic fit, project performance, machine learning, decision trees, boosting, bagging

Other note

Citation