Extract freight rate cards into a reviewed rules workspace

Why extracting the rate table alone still breaks freight bill audit

When an enterprise shipper signs or renews twenty road-transport contracts across full-truckload (FTL) and part-truckload (PTL) carriers, procurement receives twenty differently formatted PDF annexures. One transporter presents a five-page zone-to-zone grid in ₹ per kg; another lists lane rates by vehicle body type (17 ft closed, 20 ft container, 32 ft MXL) in ₹ per trip; and both bury the rules that actually drive 30% of invoice disputes inside small-print footnotes on page seven.

Those footnotes define the volumetric conversion factor (for example, 1 CFT = 8 kg versus 1 CFT = 10 kg), the minimum chargeable weight per lorry receipt (minimum 50 kg per docket), docket/FOV charges, diesel-price escalation formulas, detention free-time windows and Out-of-Delivery-Area (ODA) pincode surcharges—components reflected in public PTL tariff calculators such as the VXpress — Logistics Rate Calculator and consumed by enterprise freight audit engines like the Oracle Transportation Management — Match Rule Manager.

If a document-extraction tool scrapes the numerical grid on page two into a CSV while missing the footnote on page four that sets a ₹1,200 minimum freight per LR, every small shipment invoice will fail matching in the TMS. Conversely, if an OCR error misreads ₹6.50/kg as ₹65.0/kg and publishes it directly to the active rate master, the system could auto-approve an inflated invoice before anyone notices. Following the human-verification and source-traceability principles in the NIST — AI 600-1 Generative AI Profile, rate-card extraction must land in a Reviewed Rules Workspace where every candidate rule is linked to its source PDF bounding box and tested against sample shipments before activation.

Model a freight contract as four linked rule layers, not a flat spreadsheet

To feed a TMS rate engine or freight settlement module cleanly, configure the extraction schema to capture four distinct layers from each carrier annexure:

1. Contract header and validity metadata: Carrier legal entity, vendor code, service mode (FTL, PTL Surface, Express Air), contract reference, effective start/end dates and supersession link to the prior rate version.
2. Base lane or zone rate rows: Origin hub/zone, destination hub/zone, vehicle type or weight slab, numerical rate, currency and explicit unit of measure (PER_KG, PER_TON, PER_CASE, FLAT_TRIP).
3. Chargeable-weight and minimum modifiers: Volumetric conversion divisor/multiplier (CFT to kg or CBM to kg), minimum chargeable weight per LR, minimum base freight per LR and docket/waybill fee.
4. Conditional surcharges and accessorial clauses: Fuel surcharge formula and base diesel price reference, ODA pincode distance tiers, multi-point pickup/drop fees, detention free hours and per-day slab rates, and green-tax/toll pass-through rules.

Illustrative example: extracting 80 tariff rows and surcharge clauses from a PTL/FTL PDF annexure

Consider a procurement and freight-audit team running a pilot extraction on a multi-page PTL and FTL carrier contract PDF containing 80 distinct tariff rows and surcharge clauses. Instead of importing all 80 rows blindly, the extraction pipeline classifies each extracted candidate rule by confidence and structural clarity inside the review workspace.

The arithmetic across the 80 extracted contract items shows why staged review is indispensable:

• 62 standard lane rows (77.5%) extract cleanly with exact source-page coordinates, saving hours of manual data entry.
• 12 surcharge rules (15.0%) sit inside merged cells (diesel ratchet and ODA distance bands) and are isolated for structured formula verification.
• 6 prose footnotes (7.5%) govern volumetric conversion and minimum charges; holding these 6 items for explicit human parameterisation (62 + 12 + 6 = 80) prevents systemic freight-calculation errors across thousands of future invoices.

Run a 10-shipment simulation check before activating the rate version

Even after a procurement analyst visually checks the 80 extracted rules against the PDF, subtle interpretation differences between the carrier and the shipper can remain—such as whether the 12% fuel surcharge applies only to Base Freight or to Base Freight + ODA Charge.

Before flipping the extracted rate card from Draft Reviewed to Active for Settlement:

1. Replay 10 representative historical or synthetic shipments (including a lightweight bulky carton that triggers volumetric weight, a small parcel that hits the minimum LR charge, an ODA pincode delivery and a multi-stop FTL trip).
2. Compare the workspace's calculated freight total against the expected check-sheet total agreed with the carrier during contract negotiation.
3. Lock the approved rate version with dual sign-off (Procurement Owner + Finance/Freight Audit Owner) and an immutable link to the signed contract PDF.

Once activated, the structured tariff feeds directly into your invoice match tolerances—such as those managed in Oracle Transportation Management — Match Rule Manager—so every matched or disputed carrier invoice line cites the exact contract clause and page number from the workspace.

Next actions

1. Audit three of your most complex carrier rate-card PDFs and list every footnote, volumetric factor, minimum charge and surcharge formula that sits outside the main origin-destination grid.
2. Require any tariff-extraction tool to highlight the exact source bounding box on the contract PDF for every extracted number, unit of measure and effective date.
3. Block direct imports from OCR outputs into live TMS/ERP rate tables; enforce a Draft -> Commercial Review -> 10-Shipment Simulation -> Dual Sign-Off lifecycle.
4. Standardise a digital rate-card annexure template for upcoming carrier RFQs so new contracts arrive in both signed PDF and machine-validatable tabular format.