Corporate Financial Sanctions Intelligence Scripts

Corporate, Financial & Sanctions Intelligence Scripts

This section details the specialized Python scripts used for gathering and analyzing corporate structures, financial filings, and global sanctions lists. These tools are primarily utilized during Phase 3 (Collation & Entity Resolution) and Phase 4 (Chronological & Relational Processing) to verify identities and map financial networks.

Corporate Intelligence Aggregator

The corporate_intel.py script provides a unified interface for cross-jurisdictional company research. It aggregates data from multiple international registries to build a comprehensive profile of legal entities.

Data Source Implementation

The script implements specific wrapper classes for various global data providers:

Provider	Class Name	Authentication	Data Scope
UK Companies House	CompaniesHouse	API Key (CH_API_KEY)	GB Registration, Officers, PSC
US SEC EDGAR	SECEdgar	User-Agent only	US Public Filings (10-K, etc.), CIK
GLEIF	GLEIF	None	Global LEI records, parent entities
ICIJ	ICIJOffshoreLeaks	None	Offshore Leaks reconciliation

Key Functions and Data Flow

The primary entry point is the CorporateIntel class, which orchestrates searches across all enabled providers.

1. Entity Search: CorporateIntel.investigate(query) triggers parallel searches.
2. Standardization: Raw responses are mapped to a CompanyRecord dataclass.
3. Deep Dive: For UK entities, get_officers and get_psc (Persons with Significant Control) are called to resolve beneficial ownership.

Corporate Intelligence Architecture

This diagram maps the logical data sources to their respective Python implementation classes.

Title: Corporate Intelligence Data Flow

graph TD
    subgraph "Natural_Language_Space"
        A["Company Name / Query"]
        B["Beneficial Ownership"]
        C["Global Identifier (LEI)"]
    end

    subgraph "Code_Entity_Space"
        D["CorporateIntel.investigate()"]
        E["CompaniesHouse.search()"]
        F["SECEdgar.search()"]
        G["GLEIF.search()"]
        H["CompanyRecord (Dataclass)"]
        I["CompaniesHouse.get_psc()"]
    end

    A --> D
    D --> E
    D --> F
    D --> G
    E --> H
    F --> H
    G --> H
    B --> I
    I --> H

Loading

---

Financial Analysis & Anomaly Detection

The financial_analysis.py script performs forensic accounting on US public company filings retrieved via the SEC EDGAR API. It focuses on identifying statistical anomalies and financial distress.

Forensic Methods

The script implements three primary analytical frameworks:

• Benford's Law Test: benfords_law_test(values) checks the distribution of the first digits in financial data. It calculates a chi-squared statistic to detect potential manual manipulation.
• Altman Z-Score: altman_z_score(...) predicts bankruptcy risk by weighing five financial ratios (Working Capital, Retained Earnings, EBIT, Market Cap, and Revenue) against Total Assets.
• YoY Variance: yoy_variance(values, labels) flags year-over-year changes exceeding a 20% threshold.

Data Extraction

The FinancialAnalyser class uses get_company_facts(cik) to fetch XBRL data directly from the SEC. It then uses extract_metric_series to normalize disparate taxonomy tags (e.g., us-gaap:Revenues) into a time-series list for analysis.

---

Sanctions & PEP Screening

The sanctions_screen.py script facilitates fuzzy matching of names against global sanctions lists, including OFAC (US) and HMT (UK).

Matching Engine

The SanctionsScreener class employs a multi-tiered matching strategy to account for spelling variations and aliases:

1. Normalization: The normalise_name function strips titles (MR, SIR, etc.) and punctuation, converting strings to uppercase.
2. Exact Match: Direct comparison of normalized strings.
3. Fuzzy Match: Uses Jaro-Winkler similarity (via the jellyfish library) to generate a score between 0.0 and 1.0.
4. Phonetic Match: Generates Soundex and Metaphone keys to identify names that sound similar but are spelled differently.

List Management

The SanctionsListManager handles the lifecycle of the data:

• download_lists(): Fetches CSV data from official government URLs.
• load_all(): Parses OFAC SDN and UK HMT lists into memory-resident entries.

Sanctions Screening Logic

This diagram bridges the requirement for "Fuzzy Matching" to the specific implementation logic in the script.

Title: Sanctions Matching Pipeline

graph LR
    subgraph "Natural_Language_Space"
        Name["Subject Name"]
        List["Sanctions List (CSV)"]
    end

    subgraph "Code_Entity_Space"
        S1["normalise_name()"]
        S2["SanctionsListManager.load_all()"]
        S3["SanctionsScreener.screen()"]
        S4["jellyfish.jaro_winkler_similarity()"]
        S5["SanctionsMatch (Dataclass)"]
    end

    Name --> S1
    List --> S2
    S1 --> S3
    S2 --> S3
    S3 --> S4
    S4 --> S5

Loading

---