The Empirical Reality of IT Project Cost Overruns: Discovering A Power-Law Distribution is an actual data-driven piece of work on IT projects by Bent Flyvbjerg, Alexander Budzier, Jong Seok Lee, Mark Keil, Daniel Lunn & Dirk W. Bester. This team analysed 5,392 IT projects.
Key Observation: IT project cost overruns do not follow a normal distribution, as commonly assumed, but rather a power-law distribution, characterized by many small overruns and a fat tail with a few extreme overruns.
Sample Size and Scope: The study analyzed 5,392 IT projects completed between 2002 and 2014, with a total cost of $56.5 billion (2015 USD).
Significant Risk: The fat tail of the power-law distribution indicates that extreme cost overruns are more common than expected under a normal distribution, suggesting that IT projects are far riskier in terms of cost.
Generative Mechanism: The paper proposes that interdependencies among technological components within IT systems can lead to chain reactions, causing substantial cost overruns. A problem in one component can propagate through the system, affecting other interdependent components.
Implications for Management: Managers who assume a normal distribution of cost overruns may severely underestimate the probability of large overruns, exposing their organizations to greater risk. The study underscores the importance of realistically assessing and mitigating these risks upfront in IT project management.
Empirical Validation: Through statistical analysis, including fitting the data to a power-law distribution and conducting goodness-of-fit tests, the study provides strong evidence that IT project cost overruns indeed follow a power-law distribution.
Ruling Out Rival Explanations: The study carefully rules out alternative explanations for the observed distribution, including potential biases due to project size, type, data source, and other factors.
Broader Impact: The findings suggest a need for a paradigm shift in how IT project risks are assessed and managed, advocating for a more realistic understanding of cost overrun probabilities.