Research synthesis · Project management & project success

Project success and failure: evidence on project managers, cognitive bias and team transparency

Strang and Vajjhala show, with machine learning on project big data and controlled experiments, that IT project outcomes depend on the project manager — experience, risk competency, resistance to incentive bias and commitment — and on team transparency during procurement.

A short narrative review of research by Narasimha Rao Vajjhala and co-authors in this area. Every claim carries an APA citation that links to the publication’s own page (abstract, key findings, DOI); the full references are listed at the end. Updated .

Synthesis

Kenneth David Strang and Narasimha Rao Vajjhala have combined machine learning on project big data with controlled experiments to ask why IT projects succeed or fail. Their central finding is that outcomes depend heavily on the project manager: experience, risk competency, resistance to biased incentives and organizational commitment.

Mining big data on about 17,430 US government IT projects, Strang and Vajjhala (2023a) used random forest classification to identify seven failure indicators with 79.9% precision, 81% recall and a ROC area of 0.849; post-hoc logistic regression confirmed project manager experience, project budget and outsourced versus in-house project managers as significant predictors. Turning to successful projects, analyses of 439 US government IT projects found that linear regression best predicted project manager commitment tenure (R² = 0.238), with line of business, years of experience and team network collaboration among the strongest predictors (Strang & Vajjhala, 2025b; Vajjhala & Strang, 2024a).

Two controlled experiments isolated the role of cognitive bias. When 16 experienced US IT project managers were offered a 400% bonus to continue a simulated failing military software project, their likelihood of cancelling it fell from 3.438 to 2.438 on a 1–5 scale (p = .005, Cohen’s d = 0.826), and risk competency — not certification — was the significant protective factor (Strang & Vajjhala, 2022a). With 24 project managers facing a COVID-19 risk event, bias again lowered decision quality, and certification and competency were the only individual factors significantly related to good decisions (Strang & Vajjhala, 2022b).

Team transparency also matters. In Fintech projects, Strang and Vajjhala (2025a) found that team members’ willingness to disclose past performance evaluations during procurement significantly improved project success after controlling for budget, end-user community size and certification, raising explained variance from 9.6% to 18.8%. Vajjhala and Strang also edited a volume on risk and contingency management during COVID-19 and other crises across nations and industries (Vajjhala & Strang, 2022).

For practitioners, the evidence points to selecting and developing project managers for risk competency and experience, designing incentives that do not reward persisting with failing projects, and building transparency about past performance into team selection.

Key claims with sources

  1. Random forest analysis of about 17,430 US government IT projects identified seven failure indicators, led by project manager experience (Strang & Vajjhala, 2023a).
  2. A 400% bonus incentive significantly lowered experienced project managers’ likelihood of cancelling a failing project (Cohen’s d = 0.826) (Strang & Vajjhala, 2022a).
  3. Willingness to disclose past performance raised the explained variance in project success from 9.6% to 18.8% (Strang & Vajjhala, 2025a).

For a longer review that sets these studies against the wider literature, see the evidence overview What predicts IT project failure and success? Evidence on project managers, bias and team transparency from machine learning and experiments

The publications and what each contributes

PublicationMethodCore finding or contribution
Strang and Vajjhala (2025a)
Information
Post-positivist quantitative hypothesis testing on Fintech project data: correlation, conditioned correlation, t-tests and ordinary least squares regression with budget, end-user community size and certification as controls.Team members’ willingness to disclose their past performance evaluations during procurement significantly improved project success after controlling for budget, end-user community size and certification, roughly doubling the explained variance in project success from 9.6% to 18.8%.
Strang and Vajjhala (2025b)
International Journal of Project Organisation and Management
Exploratory machine learning study of retrospective secondary big data (439 successful US government IT projects, 1,230 features): multidimensional scaling, silhouette scores, and linear regression, random forest and SVM regression with ten-fold cross-validation.Across 439 successful US government IT projects with 1,230 features, linear regression outperformed random forest and SVM in predicting project manager commitment tenure.
Vajjhala and Strang (2024a)
WorldCIST 2024 (World Conference on Information Systems and Technologies)
Exploratory pragmatic machine learning study of a declassified U.S. government IT project repository (439 successful projects, 1,230 features) using radial analysis, multidimensional scaling and supervised regression (linear, random forest, SVM) with relief-based feature ranking.Supervised machine learning on 439 successful US government IT projects showed linear regression best predicted project manager commitment tenure (r² = 0.238), with line of business, experience, network collaboration tools and cross-industry team membership as the strongest features — supporting an association between higher commitment and project success.
Strang and Vajjhala (2023a)
International Journal of Information Technology Project Management
Pragmatic recursive mixed-methods design on U.S. Government Accountability Office big data (about 17,430 IT projects): structured data cleaning, random forest classification of project breach/failure, and post-hoc logistic regression in SPSS.Random forest analysis of about 17,430 US government IT projects identified seven failure indicators with 81% recall and a ROC area of 0.849.
Strang and Vajjhala (2022a)
International Journal of Information Technology Project Management
Online repeated-measures controlled experiment with 16 experienced U.S. IT project managers in a simulated military software project, using a PERT competency test, basic and incentive-biased decision conditions, paired t-test and repeated-measures ANOVA.A 400% bonus incentive significantly lowered 16 experienced IT project managers’ likelihood of cancelling a failing project (from 3.438 to 2.438 on a 1–5 scale; p = .005, Cohen’s d = 0.826).
Vajjhala and Strang (2022)
IGI Global (edited book)
Edited volume of 16 chapters; chapter methods include inferential statistics with auditing theory, maturity-model development, financial time-series contagion analysis, Lean Six Sigma case study, conceptual and sociological analysis, and regional empirical studies.The volume compiles evidence on risk and contingency management across nations and industries — from petroleum and automotive risk models to cybersecurity, healthcare informatics and educational inequality — offering business leaders lessons for future crises.
Strang and Vajjhala (2022b)
The Journal of Modern Project Management
Repeated-measures controlled quasi-experiment with 24 experienced U.S. project managers (competency test, basic decision, biased decision) analysed with Spearman correlation, MANOVA, repeated-measures ANOVA, Friedman test and post-hoc t-tests.In a controlled experiment with 24 experienced US project managers, a biasing incentive significantly lowered decision quality, and certification and risk competency were the only individual factors significantly related to good decisions (rho = 0.549 to 0.712).

Key terms in this area

Random forest
An ensemble machine learning technique that builds many decision trees and combines their votes to classify cases, here IT projects as failed (breached) or successful. — as used in Strang and Vajjhala (2023a)
Node purity
In random forest, the increase in node purity measures a feature's importance, which the authors liken to the change in r2 effect size when a factor is added to a regression model. — as used in Strang and Vajjhala (2023a)
ROC area (ROCa)
The area under the Receiver Operating Characteristic curve, estimating how well the model's classifications separate the two outcome classes; also called AUC. — as used in Strang and Vajjhala (2023a)
Prospect theory
A model of risk decision making in which people choose between value prospects based on loss aversion, reference points and decision weights, tending to be risk-averse for losses and risk-seeking for gains (Kahneman and Tversky, 1979). — as used in Strang and Vajjhala (2022a)
Value prospect bias
In this study, the influence of a personal economic and emotional gain (a 400% bonus) offered to a project manager for not cancelling a failing project. — as used in Strang and Vajjhala (2022a)
PERT estimate
A three-point estimate combining optimistic, most likely and pessimistic durations, used here to test project manager risk-estimation competency (correct answer: 32 days). — as used in Strang and Vajjhala (2022a)
Past performance disclosure
A team member voluntarily sharing evaluations of their performance on earlier projects when being considered for a new project team. — as used in Strang and Vajjhala (2025a)
Explained variance (R²)
The share of the variation in an outcome — here, project success — that a regression model accounts for. — as used in Strang and Vajjhala (2025a)

All key terms across the publications →

References

How to cite this synthesis

Please cite the original publications above for specific findings. To cite this overview itself:

Vajjhala, N. R. (2026, October 4). Project success and failure: evidence on project managers, cognitive bias and team transparency. Narasimha Rao Vajjhala. https://www.narasimharao.net/research/focus-areas/project-management-success-failure/

Markdown version