Overview

The Beltic Proof of Address Verification product enables organizations to verify residential and business addresses through intelligent, automated document analysis.
Built on Beltic’s AI-driven document intelligence infrastructure, this system extracts key address details, validates document authenticity, and evaluates risk signals in real time, all within a unified, compliant workflow.

It is designed to handle diverse document formats globally, such as utility bills, bank statements, tax letters, tenancy agreements, and government correspondence, ensuring precision and reliability in every verification.

Verification Flow

The Proof of Address workflow is composed of four main phases, enabling both manual and automated submission paths.

Phase 1: Document Creation

1. Create Document

• The client initiates a new proof of address record via the Beltic API.

• The system generates a document entry and corresponding file placeholders.

• Upload URLs are returned for each expected file (e.g., single-page or multi-page uploads).

2. File Upload

• Client uploads the files securely to Beltic’s cloud storage endpoints.

• File validation includes type, format, and integrity checks.

• Once all files are uploaded, they are automatically linked to the document record.

Phase 2: Automatic File Processing (Optional)

1. File Upload Detection

• The Beltic backend monitors file uploads in real time.

• When uploads are completed, file statuses are marked as “uploaded.”

• If auto-submit is enabled, the document proceeds directly to the processing stage.

Why it matters:
This flexibility allows organizations to choose between a controlled manual review or a fully automated verification flow for higher operational speed.

Phase 3: Document Submission

1. Submit Document

• The client (or system, if auto-submit is enabled) submits the document for analysis.

• Beltic validates the document metadata and queues it for AI-powered processing.

• The system supports real-time status callbacks or asynchronous polling.

Phase 4: Document Processing

1. AI-Powered Document Analysis

Beltic applies multiple AI models purpose-built for address verification:

• Optical Character Recognition (OCR): Extracts key structured fields such as:

  • Full name

  • Address lines

  • City, state, postal code

  • Country

  • Issue and expiry dates

  • Document metadata and issuer

• Document Classification: Determines the type of document (utility bill, statement, government letter, etc.) and its expected layout.

• Forgery & Tampering Detection: Detects signs of digital manipulation, editing, compression, or synthetic document generation.

• Quality Assessment: Evaluates image sharpness, contrast, and readability to ensure the extracted data meets quality thresholds.

2. Evaluation and Decisioning

Once processing is complete, Beltic’s evaluation engine:

• Assesses data completeness and classification accuracy

• Evaluates authenticity indicators and fraud signals

• Calculates a document risk score

• Determines the final status:

  • ✅ Passed - Successfully verified

  • ⚠️ Failed - Failed fraud or completeness checks

  • 🟡 Inconclusive: Unable to reach a confident decision, additional input or manual review required

  • ⏹️ Canceled - Timeout or incomplete input

3. Results Storage

• A verification record is created containing extracted data, classification confidence, and fraud indicators.

• The document record is updated with verification metadata.

• Linked sessions (e.g., onboarding or case) are updated automatically with the final status.

Phase 5: Error Handling and Recovery

Condition	Handling Strategy
Timeout (120s)	If processing exceeds 120 seconds, verification is canceled gracefully.
Partial Failure	If a processing stage fails, available results are returned and flagged as partial.
Full Failure	If both extraction and validation fail, the system marks verification as canceled.
Automatic Retry	Transient or service-related failures are retried automatically.

System Resilience:
Processing is distributed, fault-tolerant, and can continue even when partial results are available — ensuring no manual restarts are needed.

Key Capabilities

Feature	Description
AI-Powered OCR	Extracts structured address data with high accuracy across global document types.
Fraud Detection	Detects forgeries, edits, and metadata inconsistencies automatically.
Quality Assurance	Ensures documents meet readability and integrity standards before decisioning.
Configurable Automation	Supports manual submission or auto-processing modes.
Real-Time Evaluation	AI models run in parallel for fast, deterministic results.
Error Resilience	Built-in timeout, retry, and recovery logic ensures stable throughput.