Tartan’s Product Update — January 2023

Hello there,

Welcome back to our January 2023 product update. This month, we’re introducing meaningful enhancements to strengthen credit underwriting and fraud detection—helping lenders move faster while maintaining tighter risk controls.

With these updates, underwriting teams can rely on structured, machine-readable signals instead of manual document reviews, enabling more consistent and automated credit decisions.

Better Credit Decisioning with Three Underwriting Assist Scores

We’re introducing three underwriting assist scores designed to help lenders evaluate payslip submissions with higher confidence and less manual intervention. These scores work together to improve document validation, fraud detection, and data consistency checks—all through a single API.

Payslip Confidence Score

Verify document type with greater certainty

With enhanced payslip OCR capabilities, Tartan can now determine whether an uploaded document is truly a payslip. The API returns an internal confidence score indicating how likely the document is a valid payslip.

This allows clients to:

• Automatically accept genuine payslips

• Reject unrelated documents such as bank statements, Aadhaar cards, or other uploads

• Reduce manual document classification effort

Using this score, teams can design workflows that only proceed when the document meets a defined confidence threshold.

Fraud Score

Detect tampering and manipulation at multiple levels

Every payslip upload undergoes mandatory fraud analysis using a combination of metadata-level and pixel-level checks.

• Metadata Level Analysis examines file properties such as version, schema, producer keys, and creation details to identify inconsistencies.

• Pixel Level Analysis applies advanced techniques like document similarity checks and heatmap analysis to detect visual manipulation and forgery.

The resulting fraud score provides a clear signal of potential tampering, helping underwriting teams flag high-risk submissions early.

Data Mismatch Score

Identify inconsistencies between declared and extracted data

For every uploaded payslip, key details such as the employee name and employer name are extracted and matched against the information provided at the time of application or invite.

Any discrepancies—small or significant—are quantified through a mismatch score, allowing teams to:

• Detect identity or employment inconsistencies

• Route cases for manual review when required

• Automate approval or rejection decisions based on defined thresholds

One API. Multiple underwriting signals.

By combining:

• Payslip Confidence Score

• Fraud Score

• Data Mismatch Score

lenders can now build robust, automated underwriting policies using a single API integration.

These signals help reduce subjectivity, improve consistency, and scale credit decisioning without increasing operational overhead.