Analysis Methodology

How Pellucis turns property documents into forensic intelligence

Every Pellucis report is produced through a structured analysis pipeline. This page describes the methodology—what we do, why we do it in this order, how each component feeds the next, and what it means to build a complete narrative for the buyer.

Overview

Pellucis does not produce chatbot summaries. We run documents through five sequential analysis passes: text extraction and validation, financial extraction, risk extraction, governance analysis, and executive synthesis. Each pass produces structured output that feeds the next. The final report—the assessment, briefing, and narrative—is generated only after all underlying data has been extracted and validated.

The goal is a forensic intelligence briefing: not "what's wrong with this property?" but "should I buy this property, and why?" Every finding is tied to evidence in the documents. Every number is calculated, not estimated. The narrative explains how the pieces fit together.

The Five-Pass Architecture

PDF Input
  ↓
Pass 0: Text Extraction + Quality Validation
  ↓
Pass 1: Financial Extraction + Ratio Calculation
  ↓
Pass 2: Risk Extraction (severity, inversions, quality indicators)
  ↓
Pass 3: Governance Analysis (per-meeting summaries → scorecard + narrative)
  ↓
Pass 4: Executive Summary Synthesis
  ↓
Output: Report + Structured Data

Pass 0 — Text Extraction & Quality Validation

All PDFs are converted to text with page markers preserved. We validate extraction quality: sequential page numbers, real-word ratio (to catch garbled OCR), and file-level character counts. Low-confidence or failed extractions are flagged in the report header. Scanned documents use increased OCR resolution (300 DPI) for accuracy.

Pass 1 — Financial Extraction & Ratio Calculation

Revenue, expenses, surplus or deficit, reserve fund balance and adequacy, debt, arrears, special levies, insurance, and fee trajectory are extracted from financial statements, AGM reports, and meeting minutes. We calculate ratios—operating ratio (expenses ÷ revenue), reserve funding percentage, runway years, arrears rate—that contextualize every other finding. Budget overruns above 20% are flagged. Jurisdiction-specific statutory timelines (reserve fund studies: Ontario condos 3 years, Ontario co-ops 5 years, BC strata 5 years, Alberta 5 years) inform gap detection.

Pass 2 — Risk Extraction

Risks are extracted by category (Financial, Structural, Livability) and severity (Critical, High, Medium). Each finding receives a modifier: Unmitigated (no action taken), In Progress (board is addressing it), Resolved (historical record only), or Recurring (escalates severity). Pellucis also performs Munger Inversion—systematically identifying what should appear in the documents but does not. Missing reserve studies, insurance discussion, meeting chronology gaps, and property envelope assessments for aging structures are governance gaps. Quality indicators—reserve adequacy, current depreciation report, multiple quotes obtained, no litigation—are also extracted. The Kill Switch score is derived from unmitigated and recurring risks plus governance gaps; resolved risks do not count.

Pass 3 — Governance Analysis (Two-Stage)

Stage A produces structured summaries for each meeting: attendees, quorum, action items, decisions, complaints, fines, legal references, and procurement discipline. Stage B aggregates across meetings to produce the Community Health scorecard (action item resolution, meeting attendance, quorum health, response time, procurement discipline, financial compliance, conflict density) and the narrative—board character, community dynamics, and trajectory. Problem units are identified by threshold: 3+ mentions across meetings, or 2+ different issue types (e.g., noise and unpaid fines). Tribunal references (CRT for BC, CAT for Ontario) count as conflict indicators. A minimum of 4 meetings is required for full scoring; fewer triggers a confidence discount.

Pass 4 — Executive Summary Synthesis

Pass 4 receives outputs from all previous passes and produces the verdict and briefing. The decision framework is deterministic: Kill Switch ≥ 70 and Community Health ≥ 60 → Proceed; Kill Switch 40–69 with Community Health ≥ 60 → Negotiate; Kill Switch < 40 or Community Health < 60 → Caution; any single critical unmitigated risk > $100k overrides to Caution. Override flags (Underfunded, Legal Exposure, Insurance Risk) may apply. When the verdict is Negotiate, we derive a quantified exposure from identified costs. Fundamentals—reserves, surpluses, engaged ownership—can support Negotiate even with significant capital overruns; Caution is reserved for systemic compromise.

Why This Structure

The sequential pass design exists for three reasons.

1. Separation of concerns. Financial extraction requires different prompts and validation than risk extraction. Mixing them produces inferior results. Each pass is optimized for its task.
2. Downstream dependency. Pass 2 (risk) needs financial context—a $50k levy means something different when reserves are $2M vs. $200k. Pass 3 (governance) needs meeting chronology. Pass 4 (synthesis) needs all of it. The pipeline order is not arbitrary.
3. Auditability. Every output from every pass is retained. If a finding is questioned, we can trace it to the exact pass and, where applicable, the page citation. The report is not a black box.

The alternative—a single prompt that "analyzes everything"—produces generic summaries and hallucinated numbers. Pellucis extracts first, calculates second, synthesizes last.

How It All Plays Together

The passes are not isolated. They integrate in specific ways:

• Financials → Risk context. A budget overrun is a governance signal; the size of the overrun relative to reserves determines severity. Operating ratio (expenses ÷ revenue) flags deficit operations as financial risk.
• Risk → Governance gaps. Munger Inversion identifies missing documents—no reserve study, no envelope assessment for an aging property. These gaps feed the "What Pellucis Looked For and Did Not Find" section and inform the buyer's pre-closing action list.
• Governance → Community Health. Meeting summaries produce the scorecard. Trajectory compares the first half of meetings to the second—improving, stable, or declining. Problem units consume council time; conflict density (fines, tribunal references) reduces the score.
• All passes → Synthesis. Pass 4 receives financial snapshot, risk register, governance gaps, quality indicators, Kill Switch score, Community Health score, and governance narrative. It applies the decision framework and writes the briefing. The briefing reflects the data; the data is not reverse-engineered to fit a predetermined verdict.

Building the Complete Narrative

A complete narrative answers "what is this property, and what does it mean for the buyer?" in three dimensions:

Together, these three produce a coherent picture. A property with strong reserves and engaged ownership but a single capital overrun receives Negotiate with a quantified exposure—because the fundamentals support recovery. A property with deficit operations, depleted reserves, and governance dysfunction receives Caution—because the system itself is compromised. The narrative explains why.

Documents & Information Required

Pellucis analyzes the full status certificate package, strata document package, or co-op disclosure package that the seller or property manager provides. The engine is built for specific document types. What we look for—and what we do when it is missing—drives the quality of the narrative. Accuracy depends on document completeness. For how to submit your package, see Upload.

Minimum for a Valid Report

At minimum, we need financial statements and meeting minutes (AGM and/or board). Financial statements provide revenue, expenses, reserves, arrears, and the basis for every ratio. Meeting minutes provide governance signals—board decisions, procurement discipline, action items, complaints, and trajectory. Without both, the report cannot produce a complete narrative.

For full governance scoring and trajectory analysis, we require a minimum of 4 meetings. Fewer meetings triggers a confidence discount on Community Health and board character—the narrative will note the limited data window.

Documents We Review

Pellucis accepts and extracts from the following document types. Include everything the seller or property manager provided—the more complete the package, the more thorough the analysis.

Statutory Timelines (Governance Gap Detection)

Pellucis flags missing documents against jurisdiction-specific requirements. A missing study within the statutory window is an inversion.

• Ontario condo: Reserve fund study every 3 years
• Ontario co-op: Reserve fund study every 5 years
• BC strata (5+ units): Depreciation report every 5 years (mandatory as of July 2024)
• Alberta: Reserve fund study every 5 years

When Documents Are Missing

Pellucis performs Munger Inversion: we systematically identify what should appear in the documents but does not. A missing reserve study within the statutory timeline (see above) is a governance gap. No insurance discussion in meeting minutes for an aging property is a gap. Missing property envelope assessment for aging construction (e.g., 1980-era) is a gap. Each inversion is listed in "What Pellucis Looked For and Did Not Find" with a recommended action for the buyer before closing.

We do not fill gaps with speculation. If a field cannot be extracted—no reserve balance, no meeting minutes—we set it to null and flag it in the data gaps. The report states what was found and what was not. The narrative reflects the available evidence.

Evidence & Citations

Every finding in a Pellucis report is tied to the source documents. Numerical values—reserve balance, operating ratio, arrears rate, runway years—are extracted from the documents and calculated, not estimated. The narrative is written by AI. The numbers are not. Every assessment, flag, and score is validated against the source data before delivery.

Risk findings include the exact quote or sentence that evidences the risk, plus the file and page citation. Governance gaps specify what is missing and why it matters for this property at this age. Dimension ratings (Strong, Adequate, Monitoring, Concerning, Critical) are derived from the extracted data and the decision framework—not reverse-engineered to fit a predetermined verdict.

The engine uses page markers (--- [FILE: filename | PAGE: number] ---) to preserve traceability. Every output from every pass is retained. When a client questions a finding, we can trace it to the exact pass and, where applicable, the page citation. The report is not a black box.

When documents are incomplete—missing meeting minutes, no financial statements—we flag data gaps. The report states what was found and what was not. We do not fill gaps with speculation. Pellucis produces a briefing, not a pitch. We state. We do not suggest. The voice is direct, firm, and evidence-based. Every section earns its place in the narrative.

Related