Intermodal Drayage Demurrage Auditing: Last Free Day (LFD) & Port Dispute Engines

Core Architecture of an Automated LFD & Dispute Engine

A production-grade demurrage audit system is built around four tightly integrated subsystems:

1. Real-Time Event Ingestion Layer

The foundation is an event-driven ingestion pipeline that consumes terminal EDI messages (typically 214, 322, and custom terminal status feeds), gate appointment confirmations, and truck GPS/ELD pings. Strong Tower recommends a normalized event store using a schema similar to the following (PostgreSQL example):

CREATE TABLE container_events (
    event_id UUID PRIMARY KEY,
    container_number VARCHAR(20) NOT NULL,
    booking_number VARCHAR(30),
    event_type VARCHAR(50) NOT NULL, -- 'AVAILABLE', 'GATE_CLOSED', 'APPT_SCHEDULED', 'APPT_REJECTED', 'CUSTOMS_HOLD', 'RELEASED'
    event_timestamp TIMESTAMPTZ NOT NULL,
    terminal_id VARCHAR(20),
    source_system VARCHAR(30),
    raw_payload JSONB,
    processed BOOLEAN DEFAULT FALSE,
    created_at TIMESTAMPTZ DEFAULT NOW()
);

CREATE INDEX idx_container_events_container_time ON container_events(container_number, event_timestamp);
CREATE INDEX idx_container_events_type ON container_events(event_type);

This table serves as the immutable system of record for all LFD calculations. Downstream processors subscribe to new events via logical replication or a message queue (Kafka, RabbitMQ, or AWS SNS/SQS).

2. Last Free Day Calculation Engine

The LFD engine applies terminal-specific business rules, OSRA-22 adjustments, and calendar exceptions. Key logic includes:

• Base free days per terminal tariff (pulled from a rate matrix table linked via /practices/carrier-rate-matrix-optimization).
• Automatic deduction or extension for documented terminal closure days (weekends, holidays, force majeure, or published gate closures).
• Freeze of the LFD clock during customs or FDA holds when evidence exists in the event store.
• Adjustment for failed gate appointments caused by terminal system outages or dual-transaction lockouts.

The calculation is idempotent and versioned. Every time new evidence arrives, the system recalculates the effective LFD and stores the result with an audit trail:

CREATE TABLE lfd_calculations (
    calculation_id UUID PRIMARY KEY,
    container_number VARCHAR(20) NOT NULL,
    base_lfd_date DATE,
    adjusted_lfd_date DATE NOT NULL,
    calculation_version INT NOT NULL,
    rules_applied JSONB NOT NULL,
    evidence_event_ids UUID[] NOT NULL,
    calculated_at TIMESTAMPTZ DEFAULT NOW()
);