StakeNet: Solving the "Missing Stakeholder" Problem via Social Networks

TL;DR

Project failure is often not a technical problem, but a sociological one: we simply forget to talk to the right people. StakeNet transforms stakeholder analysis from a subjective manual task into a quantitative science. By treating stakeholders as nodes in a social network and using recommendation-based "snowballing," it identifies 50% more critical roles than traditional checklists and provides a mathematically grounded priority list.

The "Invisible" Stakeholder: A $20 Million Mistake

In large-scale systems, the stakeholder landscape is a labyrinth. The authors point out a staggering reality: overlooking stakeholders is perhaps the most common mistake in software development. In one cited case, a project was cancelled at a cost of $20 million because a single key stakeholder was ignored.

Existing methods (checklists, generic roles, or semi-structured interviews) suffer from local bias. A manager only sees their department; a developer only sees their technical peers. No one has the "God's eye view"—until now.

Methodology: Crowdsourcing Project Sociology

StakeNet moves away from the "expert-led" model to a collaborative model. The core workflow is elegant:

1. Snowballing Recommendations: Start with a small seed group. Ask them: "Who else matters, and how much (salience)?"
2. Network Construction: Build a directed graph where an edge from A to B means "A recommends B."
3. Algorithmic Prioritization: Instead of guessing who is important, StakeNet runs social network measures:

Why Centrality Matters

The paper utilizes several key metrics to find the "real" power players:

• Betweenness Centrality: Finds "brokers"—people who bridge different departments.
• PageRank: Similar to Google search, it finds stakeholders who are recommended by other highly-recommended people.
• In-degree: A simple popularity contest—who does everyone think we should talk to?

Empirical Evidence: The RALIC Case Study

The authors tested StakeNet on the RALIC project at University College London—a complex system replacing access control for 30,000 users.

Key Performance Metrics

Comparing StakeNet against the existing project management "Checklist" method yielded dramatic results:

• Recall: StakeNet found 90% of actual stakeholders, whereas the traditional method found only 40%.
• Precision: Despite finding more people, StakeNet remained highly accurate (90% precision).
• The "Killer" Discovery: StakeNet identified "External Library Users" as a high-salience group. The original project team had ignored them, only to realize years later that issuing cards to 17,000 external members was financially impossible, forcing a major change in project scope.

Deep Insight: Beyond the Org Chart

One of the most profound findings is that StakeNet outperforms individual experts. Stakeholders’ individual prioritisations had an average accuracy ($\rho$) of only 0.360, while the collective network reached 0.785+.

Why? Because individuals are biased. Developers prioritize developers. Managers prioritize managers. StakeNet’s network topology breaks these silos by looking at the global flow of recommendations. It even identifies "bottlenecks"—people with high influence but low involvement—alerting project managers to potential communication risks before they become crises.

Conclusion & Future Outlook

StakeNet proves that the "Social Network" of a project is as important as its source code. By applying graph theory to requirements engineering, the authors provide a scalable, objective, and highly accurate way to ensure no critical voice is left unheard.

Future Work: The authors are now looking at "Dynamic StakeNet"—how to track shifting stakeholder power over time as a project moves from requirements to maintenance.

Final Takeaway

If you are managing a large-scale project, stop relying on your gut or a static checklist. Build the network, run the algorithms, and let the "wisdom of the crowd" guide your requirements.