Product Vision & Roadmap¶

Vision Statement¶

CalcBridge exists to eliminate the friction between complex financial analysis and timely decision-making. We envision a world where:

• Compliance testing is instantaneous, not a multi-hour ordeal
• Trade decisions are backed by real-time impact simulation
• Every calculation has complete audit provenance
• Data reconciliation happens automatically with intelligent exception management
• Teams collaborate seamlessly across geographies and time zones
• Systems are resilient, self-healing, and transparent

Problem Statement¶

The Current State of CLO Analytics¶

CLO portfolio management relies heavily on Excel workbooks that have evolved over decades. These workbooks contain critical business logic but suffer from fundamental limitations that create significant operational risk and inefficiency.

Pain Point 1: Compliance Testing Hell¶

User Voice:

"Every month-end, I spend two full days just running compliance tests. Half my time is double-checking my own work because the stakes are so high." - Senior CLO Analyst

CalcBridge Solution:

• Automated compliance test execution on every data change
• Real-time pass/fail dashboard with drill-down capability
• Historical compliance trend analysis
• Configurable alert thresholds for approaching limits
• Predictive compliance warnings before breaches occur

Pain Point 2: Trade Decision Paralysis¶

User Voice:

"When a good trade opportunity comes up, I need to know within minutes if it works for my portfolio. Instead, I'm copying numbers into Excel for half an hour while the window closes." - Portfolio Manager

CalcBridge Solution:

• What-if scenario builder with instant calculation
• Side-by-side comparison of base vs. scenario metrics
• Automatic compliance test re-evaluation for proposed trades
• Multi-trade scenarios with cumulative impact analysis
• Scenario sharing and collaboration tools
• Export scenarios for audit documentation

Pain Point 3: Audit Nightmares¶

User Voice:

"Our auditors asked for evidence of calculation changes over the past year. It took us three weeks to piece together what we could from email attachments and file shares." - Compliance Officer

CalcBridge Solution:

• Complete audit trail of every data change and calculation
• User attribution with timestamp for all operations
• Version comparison and rollback capability
• SOC 2 compliant logging and retention (7 years)
• Export audit evidence on demand

Pain Point 4: Data Reconciliation Burden¶

User Voice:

"Every month I spend a full day comparing our numbers to the trustee report. Half the variances are the same issues we had last month, but we have no systematic way to track them." - Data Operations Specialist

CalcBridge Solution:

• Automated position matching across data sources
• Variance detection with configurable tolerances
• Exception workflow with resolution tracking
• Reconciliation history with trend analysis
• Auto-resolution rules for known variance patterns

Pain Point 5: Operational Fragility¶

User Voice:

"We discovered our WARF calculation had been wrong for three months because someone accidentally overwrote a formula. There was no way to know until an investor questioned the numbers." - Risk Manager

CalcBridge Solution:

• Dead letter queue for failed operations with alerting
• Circuit breaker patterns for external dependencies
• Comprehensive health monitoring and metrics
• Formula validation before evaluation
• Data quality scoring and anomaly detection

Target Audience¶

Primary Personas¶

1. CLO Analyst¶

Key Jobs to Be Done:

• Run compliance tests before trustee reports
• Analyze portfolio composition and trends
• Generate ad-hoc reports for stakeholders
• Investigate data discrepancies
• Review reconciliation exceptions

2. Portfolio Manager¶

Key Jobs to Be Done:

• Evaluate potential trades before execution
• Monitor portfolio health and compliance cushion
• Present portfolio strategy to investors
• Collaborate with analysts on optimization
• Stress test portfolio under various scenarios

3. Compliance Officer¶

Key Jobs to Be Done:

• Review compliance test results
• Generate audit evidence on demand
• Monitor for approaching covenant limits
• Investigate and document exceptions
• Manage reconciliation variance resolution

4. Data Operations Specialist¶

Key Jobs to Be Done:

• Upload and validate source workbooks
• Configure column mappings for new sources
• Troubleshoot data quality issues
• Maintain data dictionary and documentation
• Monitor schema drift and mapping health

5. Risk Manager¶

Key Jobs to Be Done:

• Run stress test scenarios across portfolios
• Analyze concentration risk and exposure
• Monitor compliance cushion trends
• Evaluate rating migration impact
• Generate risk reports for stakeholders

6. System Administrator¶

Key Jobs to Be Done:

• Manage tenant configuration
• Provision and deprovision users
• Configure role-based access
• Monitor system health and usage
• Respond to operational incidents
• Review DLQ and retry failed operations

7. Developer / Integrator¶

Key Jobs to Be Done:

• Build data pipelines from upstream systems
• Create custom reporting dashboards
• Automate routine operations
• Integrate with trading systems
• Consume webhook events for downstream processing

Product Roadmap¶

Phase 1: Core Platform (Complete)¶

Delivered Value:

• 99.7% upload success rate
• < 200ms average API response time
• Zero cross-tenant data access incidents
• 50+ Excel functions supported
• 12+ servicer alias profiles

Phase 2: Advanced Analytics (Complete)¶

Expected Value:

• 80% reduction in compliance testing time (4-8 hours → 30 minutes)
• < 3 second what-if calculation time
• 100% audit trail coverage
• 99.5% reconciliation variance detection accuracy
• Zero manual covenant verification

Phase 3: Real-time Collaboration (Planned)¶

Expected Value:

• 50% reduction in email communication
• Real-time team visibility into portfolio changes
• Structured approval workflows for trades
• Mobile accessibility for alerts
• Shared exception management

Phase 4: AI-Powered Indenture Intelligence (Future)¶

The End-to-End Pipeline Vision¶

graph LR
    A["Indenture PDF/DOCX"] --> B["AI Document Parser"]
    B --> C["Extracted Data Points<br/>(OC/IC tests, concentration<br/>limits, eligibility criteria)"]
    C --> D["Human-in-the-Loop<br/>Verification"]
    D --> E["Auto-Configure<br/>Compliance Tests"]
    E --> F["CalcBridge<br/>Test Engine"]
    F --> G["Trustee Reports<br/>& Evidence Packages"]

Why This Is Tractable¶

CLO indentures are semi-standardised documents. While each deal has unique terms, the structure and vocabulary are consistent across the market. This makes them an ideal candidate for domain-specific AI extraction:

• Predictable sections: OC/IC test definitions, concentration limits, eligibility criteria, waterfall provisions
• Consistent terminology: Industry-standard terms (WARF, WAS, WAL, Diversity Score) appear across nearly all deals
• Finite data point set: Starting with 20-50 key data points per deal covers the vast majority of compliance test configuration needs
• High-value automation: Manual test configuration from indentures currently takes days per deal; AI extraction reduces this to minutes with human verification

Technical Approach¶

• RAG-based extraction: Retrieval-Augmented Generation with domain-specific pipelines tuned for CLO legal documents
• Domain-specific models: Fine-tuned on CLO indenture corpus for high-accuracy extraction of financial covenants, test thresholds, and eligibility criteria
• Structured output: Extracted data points map directly to CalcBridge compliance test configuration schema
• Confidence scoring: Every extracted data point includes a confidence score; low-confidence items are flagged for human review
• Natural language querying: Same RAG infrastructure enables conversational querying of deal documents (e.g., "What is the OC trigger for Class A notes?")

Human-in-the-Loop Verification¶

AI extraction is never fully autonomous. Every extracted data point passes through human verification:

1. Extraction: AI parses indenture and extracts structured data points with confidence scores
2. Review UI: Analyst reviews extracted values against source document with highlighted provenance
3. Confirmation: Analyst confirms, corrects, or rejects each data point
4. Audit trail: Every confirmation/correction is logged with user attribution and timestamp
5. Learning loop: Corrections feed back into model improvement over time

Competitive Context¶

Staffing Requirements¶

• 2-3 ML engineers with NLP/document extraction experience
• Domain expertise in CLO/structured credit (existing team provides this)
• Phased hiring aligned with pipeline development milestones

Roadmap Detail¶

Near-term (representational):

Medium-term (representational):

Expected Value:

• End-to-end automation: From indenture document to configured compliance tests in minutes, not days
• Proactive risk identification: Anomaly detection before covenant breaches occur
• Democratized document access: Non-technical stakeholders can query deal terms in natural language
• Reduced onboarding friction: New deals configured automatically from their governing documents
• Audit-grade provenance: Every AI-extracted data point traceable to source document with human verification stamp
• Continuous learning: Model accuracy improves with each human-verified correction

Success Metrics & KPIs¶

Business Metrics¶

Technical Metrics¶

User Engagement Metrics¶

Competitive Landscape¶

Market Position¶

Competitive Advantages¶

1. Purpose-Built for CLO - Not a generic tool adapted for finance
2. Excel-Native Mental Model - Users work with familiar concepts
3. Compliance-First Design - Audit trail and testing are core, not add-ons
4. What-If Simplicity - Complex simulation made accessible
5. Multi-Tenant Economics - Enterprise features at accessible pricing
6. Reconciliation Built-In - Data quality is core workflow
7. Operational Resilience - Self-healing with visibility

Risk Assessment¶

Implementation Principles¶

Design Principles¶

1. Excel Familiarity - Preserve mental models from Excel workflows
2. Compliance First - Every feature considers audit and compliance impact
3. Instant Feedback - Calculations should feel immediate
4. Safe by Default - No silent failures, explicit error handling
5. Progressive Disclosure - Simple by default, powerful when needed

Technical Principles¶

1. Safe Calculation - AST-based formula evaluation, never unsafe execution
2. Vectorized Performance - Pandas/NumPy for 100x speed over loops
3. Tenant Isolation - Row-Level Security for every query
4. Observable Operations - Metrics, tracing, and logging by default
5. Resilient Architecture - Circuit breakers, DLQ, graceful degradation

Data Principles¶

1. JSONB Flexibility - Schema evolution without migrations
2. Audit Everything - Before/after logging for all changes
3. Quality by Default - Validation and anomaly detection
4. Reconciliation Ready - Position matching and variance tracking

Last Updated: 2026-03-22 | Version 2.1.0