A Replicable Architecture for Institutional Transparency
Applied Coherence Institute (ACI) & Sovereign Integrity Institute (SII)
Authors: Nathan Veil (ACI) & David Humble (SII)
Date: June 1, 2026
Status: Technical Report – Open Architecture
License: CC BY-NC 4.0
Abstract
Markets and societies lack continuous, entity‑level, verifiable signals for governance risk. Existing mechanisms (credit ratings, ESG scores, corruption indices) are infrequent, self‑reported, or country‑level. This paper describes the design, implementation, and early validation of the Coherence Oracle – a live, public, real‑time governance risk assessment system for corporate and government entities. The Oracle provides weekly coherence scores (0–100) for over 200 entities across six jurisdictions, using publicly available data and a transparent, weighted pillar methodology. The paper defines coherence operationally, grounds extraction risk in existing governance literature, describes the scoring framework, reports early pilot validation, and provides a replicable architecture for other jurisdictions. The Oracle is positioned as a next‑generation transparency instrument, not as a protest or campaign.
Keywords: governance risk, transparency, institutional integrity, coherence, oracle, accountability
1. Introduction
Governance risk – the risk that an entity will fail to meet its stated obligations due to corruption, opacity, or regulatory capture – is systematically underpriced in credit, insurance, and procurement markets (Klitgaard, 1998; Kaufmann et al., 2010). Credit ratings focus on financial solvency, not governance integrity. ESG scores rely on self‑reporting (Berg et al., 2022). Corruption indices are country‑level and annual (Transparency International, 2025). As a result, governance risk remains invisible until a scandal, enforcement action, or collapse reveals the hidden liability.
This paper describes the Coherence Oracle, a live governance risk assessment system that addresses this gap. The Oracle provides weekly, entity‑level coherence scores for corporations and government entities across six jurisdictions (Thailand, Hong Kong, Singapore, the United States, China, and Laos), using only publicly available, verifiable data.
The paper makes three contributions:
- It operationalizes coherence as a measurable governance construct.
- It presents a replicable architecture for entity‑level governance risk scoring.
- It reports early validation results from a pilot study of 199 entities.
The Oracle is not a proprietary black box. Its methodology is fully published. Its code is open for inspection (with implementation‑specific parameters omitted to protect ongoing refinement). The goal is to provide a blueprint that other sovereign witnesses can adapt to their own jurisdictions.
2. Theoretical Framework
2.1 Defining Coherence Operationally
In this paper, organizational coherence is defined as the degree of alignment between an entity’s stated governance obligations and its observable public behavior. This alignment is measured across multiple domains:
| Domain | Operationalized As |
|---|---|
| Governance | Ownership transparency, board independence, whistleblower protection |
| Legal compliance | Fines, sanctions, enforcement actions, obstruction history |
| Supply chain integrity | Forced labor indicators, wage theft, ethical sourcing |
| Tax transparency | Country‑by‑country reporting, tax haven use |
| Environmental compliance | Fines, land conflicts, resource depletion |
For government entities, additional domains are included: procurement integrity, regulatory capture, financial transparency, anti‑corruption enforcement, whistleblower protection, and access to information (see Section 4.2). Coherence scores range from 0 (lowest coherence, highest governance risk) to 100 (highest coherence, lowest governance risk).
2.2 Governance Risk and Extraction
The concept of extraction – the systematic transfer of value without corresponding regeneration – is grounded in existing governance literature. Klitgaard’s (1998) formulation of corruption as Corruption = Monopoly + Discretion – Accountability captures the structural conditions under which extraction occurs. The Worldwide Governance Indicators (Kaufmann et al., 2010) measure voice and accountability, political stability, government effectiveness, regulatory quality, rule of law, and control of corruption. The Oracle operationalizes these constructs at the entity level.
Within the Oracle framework, the term extraction risk is used as an umbrella for governance risk, corruption exposure, institutional opacity, and accountability deficits.
3. Research Questions
The Oracle is designed to enable empirical testing of three research questions:
| RQ | Question |
|---|---|
| RQ1 | Can publicly available governance indicators be combined into a stable, entity‑level coherence score? |
| RQ2 | Do low coherence scores predict future governance‑related enforcement actions? |
| RQ3 | Do coherence scores correlate with audit findings and procurement irregularities? |
These questions are not answered definitively in this paper. The paper reports early pilot validation (Section 9) and establishes a framework for longitudinal testing.
4. Scoring Framework
4.1 Corporate Pillars
| Pillar | Weight | Description |
|---|---|---|
| Governance | 25% | Ownership transparency, board independence, whistleblower protection |
| Legal & Regulatory Compliance | 25% | Fines, sanctions, debarment, obstruction history |
| Supply Chain & Labor Integrity | 20% | Forced labor, wage theft, ethical sourcing |
| Tax & Financial Transparency | 15% | Country‑by‑country reporting, tax haven use |
| Environmental Extraction | 5% | Environmental fines, land conflicts, resource depletion |
4.2 Government Pillars
| Pillar | Weight | Description |
|---|---|---|
| Procurement Integrity | 25% | Sole‑source contracts, bid rigging, documented waste/fraud |
| Regulatory Capture | 20% | Revolving door, undisclosed industry meetings |
| Financial Transparency | 20% | Audit findings, budget disclosure, unexplained expenditures |
| Anti‑Corruption Enforcement | 15% | Investigations, sanctions, prosecutions |
| Whistleblower Protection | 10% | Legal framework, documented retaliation |
| Access to Information | 10% | FOI response time, appeal success rate |
4.3 Tier Thresholds
| Tier | Score Range | Interpretation |
|---|---|---|
| Gold | 85–100 | High coherence; low governance risk |
| Silver | 70–84 | Moderate coherence; moderate governance risk |
| Bronze | 50–69 | Low coherence; elevated governance risk |
| Unrated | <50 or insufficient data | Very low coherence or insufficient data for reliable scoring |
For entities with insufficient data, a “Limited Data” flag is applied. Low scores in such cases may reflect opacity as much as extraction.
5. Data Sources
The Oracle relies exclusively on publicly available, verifiable data sources. Examples by jurisdiction include:
| Jurisdiction | Corporate Sources | Government Sources |
|---|---|---|
| Thailand | SET filings, DBD, OCCRP | NACC, SAO, OIC, G‑Procurement |
| Hong Kong | HKEX, Companies Registry | ICAC, Audit Commission, FOI logs |
| Singapore | SGX, ACRA | CPIB, AGO, EDB, FOI logs |
| United States | SEC EDGAR, DOJ, EPA | FPDS, USAspending, GAO, OpenSecrets |
| China | HKEX (H‑shares), OFAC, OCCRP | CCDI (limited), World Bank sanctions |
| Laos | OCCRP, World Bank (limited) | OCCRP, FATF grey list findings |
The methodology is jurisdiction‑agnostic. New countries can be added by extending data ingestion pipelines.
6. Technical Architecture
| Component | Technology | Purpose |
|---|---|---|
| Data ingestion | Python (Scrapy, Pandas) | Weekly scraping of public data sources |
| Scoring engine | Python + Pandas | Pillar score calculation, coherence score aggregation |
| Database | PostgreSQL | Entity data, historical scores, user accounts |
| Backend API | Node.js / Express | Serve scores to dashboard and premium API |
| Frontend dashboard | React / Next.js | Public ranking table, filters, detail pages |
| Authentication | Base44 native | Email/password login for premium users |
| Payments | Stripe | Subscription processing ($9/month, $90/year) |
| Hosting | Vercel / Base44 | Publicly accessible, no login for free tier |
The public methodology intentionally omits implementation‑specific parameters (exact data source endpoints, scoring weights micro‑adjustments, code optimizations) that are subject to ongoing refinement.
7. Freemium Model
| Tier | Price | Access |
|---|---|---|
| Free | $0 | Top 5 highest‑scoring entities (all jurisdictions) |
| Premium | $9/month or $90/year | All 199+ entities, full historical data, CSV export, API access, email alerts |
| Voucher | $0 (for journalists, researchers, witnesses) | Full premium access, subsidized by paid subscriptions |
The voucher system ensures that governance researchers, investigative journalists, and sovereign witnesses are not turned away.
8. Validation Framework
8.1 Internal Validation
| Method | Description |
|---|---|
| Inter‑rater reliability | Three independent reviewers scored a subset of 20 entities. Agreement was measured using Fleiss’ kappa (κ = 0.82, substantial agreement). |
| Sensitivity testing | Pillar weights were adjusted ±10% across 100 iterations. Entity rankings remained stable (rank correlation ρ > 0.90). |
8.2 Historical Validation
A preliminary retrospective analysis compared coherence scores for 20 entities with known subsequent governance enforcement actions (fines, sanctions, prosecutions) occurring 6–18 months after the score was calculated. The mean coherence score for entities with subsequent actions was 47 (SD = 12), compared to 78 (SD = 14) for entities without known actions. This difference is statistically significant (t(18) = 5.2, p < .001).
Limitation: This is a post‑hoc analysis. Prospective validation is ongoing.
9. Pilot Case Study: 199 Entities Across Six Jurisdictions
9.1 Sample
| Jurisdiction | Corporate Entities | Government Entities |
|---|---|---|
| Thailand | 20 | 20 |
| Hong Kong | 15 | 15 |
| Singapore | 16 | 15 |
| United States | 23 | 20 |
| China | 16 | 15 |
| Laos | 12 | 12 |
| Total | 102 | 97 |
Entities were selected based on economic significance, public data availability, and documented governance risk indicators.
9.2 Results
| Jurisdiction | Average Coherence Score (Corporate) | Average Coherence Score (Government) |
|---|---|---|
| Thailand | 71 (Silver) | 58 (Bronze) |
| Hong Kong | 68 (Silver) | 62 (Silver) |
| Singapore | 73 (Silver) | 69 (Silver) |
| United States | 65 (Silver) | 55 (Bronze) |
| China | 42 (Unrated, Data Limited) | 38 (Unrated, Data Limited) |
| Laos | 38 (Unrated, Data Limited) | 35 (Unrated, Data Limited) |
9.3 Observations
- Thai banking sector averaged Silver/Gold, consistent with regulatory pressure.
- US Department of Defense scored Unrated (opacity, limited procurement data).
- Chinese and Lao entities uniformly scored Unrated with Data Limited flag. Low scores in these jurisdictions reflect data scarcity as much as governance risk.
- Government entities consistently scored lower than corporate entities in the same jurisdiction, consistent with literature on government transparency lagging corporate disclosure.
10. Limitations
10.1 Methodological Limitations
| Limitation | Mitigation |
|---|---|
| Public data bias | Entities with more disclosures may appear worse. Flag for data scarcity. |
| Jurisdictional bias | Countries differ in reporting quality. Scores are not directly comparable across jurisdictions without adjustment. |
| Reporting lag | Public enforcement data often trails misconduct by years. Scores are directional, not real‑time. |
| Survivorship bias | Large entities leave more public footprints. The sample overrepresents large, regulated entities. |
| Validation scope | Historical validation is retrospective. Prospective validation is ongoing. |
10.2 Technical Limitations
| Limitation | Mitigation |
|---|---|
| Data scarcity (Laos, China) | “Limited Data” flag; persistent disclaimer |
| Scoring frequency weekly (not real‑time) | Sufficient for governance monitoring; real‑time not feasible with public data |
| No independent audit of methodology yet | Methodology published; third‑party audit planned |
11. Future Work
| Phase | Timeline | Activities |
|---|---|---|
| Phase 3 | Q3 2026 | Expand corporate US to 100+ companies; add Vietnam, Malaysia, Indonesia |
| Phase 4 | Q4 2026 | Government EU coverage; real‑time API alerts; independent academic validation |
| Phase 5 | 2027 | Prospective validation study correlating coherence scores with subsequent enforcement actions; insurance pilot |
12. Conclusion
The Coherence Oracle provides a live, replicable governance risk assessment system for corporate and government entities. It operationalizes coherence, grounds extraction risk in existing literature, reports early validation results, and offers a freemium model that subsidizes witness access through farm subscriptions.
The Oracle is not a protest. It is not a campaign. It is a mirror – continuous, entity‑level, and jurisdiction‑agnostic.
“The oracle does not judge. It reflects. And when the market sees the reflection, it will act.”
13. References
- Berg, F., Kölbel, J. F., & Rigobon, R. (2022). Aggregate confusion: The divergence of ESG ratings. Review of Finance, 26(6), 1315–1344.
- Kaufmann, D., Kraay, A., & Mastruzzi, M. (2010). The Worldwide Governance Indicators: Methodology and analytical issues. World Bank Policy Research Working Paper No. 5430.
- Klitgaard, R. (1998). Controlling corruption. University of California Press.
- Transparency International. (2025). Corruption Perceptions Index 2025.
- Veil, N., & Dauch, L. (2026). The Coherence Stack: From Individual Practice to Market Mirror. ACI/SII Implementation Report.
- World Bank. (2025). Enterprise Surveys & Sanctions Lists.
Correspondence: Nathan Veil, Applied Coherence Institute. consulting@appliedcoherenceinstitute.org
Leave a Reply