An enterprise-grade public reference implementation for seeing, explaining, and changing cross-team delivery exposure.
Dependency Intelligence is positioned as an operating model for senior/principal technology program and engineering delivery leaders—not as a task tracker. It turns fragmented commitments into an executive view of critical path, ownership, provider concentration, downstream blast radius, and the intervention most likely to protect the outcome.
Public reference implementation based on enterprise operating patterns. Every organization, program, person, service, commitment, and metric in this repository is fictional and synthetic.
Which cross-team dependencies threaten delivery, why, and what intervention changes the outcome?
The application provides:
- Executive dependency exposure and leadership posture
- A visible critical chain with propagation risk
- A ranked dependency register with explainable scores
- Provider concentration and commitment-health intelligence
- Historical exposure trends
- An evidence-grounded executive brief with a no-key deterministic mode
Most dependency logs describe the work. This implementation supports a decision. It separates calculated exposure from hard escalation rules, shows every scoring driver, assigns the next action and owner, and makes systemic constraints visible across programs.
Each dependency receives a 0–100 exposure score (higher is worse) from seven weighted signals:
| Signal | Weight | Leadership meaning |
|---|---|---|
| Criticality | 23% | Consequence if the commitment fails |
| Commitment health | 23% | Whether the provider is on track, at risk, uncommitted, or past due |
| Schedule proximity | 15% | Time remaining to intervene |
| Delivery confidence gap | 14% | Uncertainty in meeting the commitment |
| Downstream blast radius | 12% | Number of outcomes affected |
| Evidence staleness | 7% | Reliability of the operating signal |
| Accountability gap | 6% | Whether a named owner exists |
Explainable rules then apply: a critical past-due dependency cannot score below 85; an unowned high-impact dependency cannot score below 80; completed dependencies are capped at 10. See the full methodology.
flowchart LR
S["Synthetic JSON source"] --> M["Typed exposure model"]
M --> U["Next.js decision interface"]
U --> B["Deterministic executive brief"]
T["Future enterprise connectors"] -.-> M
A["Optional approved AI provider"] -.-> B
U --> V["Vercel / public hosting"]
The initial implementation is intentionally static and privacy-safe. No backend is required for the V1 demonstration. The domain model is ready for later Jira, Confluence, CI/CD, service-catalog, and test-management adapters.
Requirements: Node.js 22 and pnpm 10.
pnpm install
pnpm devThen open http://localhost:3000.
Validate the repository with:
pnpm checkapp/ Interactive decision interface
data/ Synthetic portfolio and dependency data
lib/ Explainable exposure methodology
docs/ Product, architecture, model, and delivery artifacts
tests/ Rendered-output checks
.github/workflows/ Continuous integration
v0.1— Executive V1 reference implementationv0.2— What-if intervention simulator and dependency scenariosv0.3— Adapter contracts for Jira, service catalogs, and CI/CD evidencev0.4— Dependency graph analytics, cycle detection, and forecast rangesv1.0— Configurable enterprise operating model with governed AI narratives
This project is designed to sit under the future Engineering Intelligence Lab umbrella alongside Release Intelligence, Engineering Control Tower, AI SDLC Governance, and Cyber Risk Command Center.
MIT licensed. See LICENSE. Do not use the synthetic scoring model as the sole basis for a real production or personnel decision without validating it against your organization’s governance model.