Release notes

What's new?

Detect

New Detect signals

We have introduced several new signals to enhance the accuracy of our tampering detection and document analysis:

• txn_data_unavailable: This signal is triggered when a bank statement is processed and no transactions are found. It indicates a lack of available transaction information for analysis.
• future_date: The future date signal identifies dates in the document that are beyond the current date.
• invalid_date: This signal detects dates in the document that are invalid or don't align with the expected period.
• future_year: The future year signal pinpoints a year in the document that is beyond the current year.
• invalid_year: The invalid year signal identifies years in the document that are invalid or do not align with the expected period.

These signals are relevant to bank statements, pay stubs, and W-2s. They will have a crucial role in upholding the integrity of document analysis, guaranteeing precise and dependable outcomes. For more detailed information, see bank statement tampering signals, pay stub tampering signals, w-2 tampering signals, and supporting data values sections.

### Detect Severity

We've introduced the new Detect Severity feature, which consolidates the detected signals within ePDF bank statements and generates an overall score indicating the likelihood of document tampering. By assessing the signals present on a bank statement, considering their severity and our confidence level in the findings, the document is assigned a severity category as High, Medium, or Low. If no signals are identified on the bank statement, the severity category remains empty i.e., N/A. This score serves as a collective representation of all individual signals and can aid in determining potential fraud when evaluating an application.

Please note that the Detect Severity is currently exclusively available through the API.

To obtain additional information about this severity mechanism, see Detect, Book-level fraud signal, Document-level fraud signal, and related FAQs.

This feature is rolled back. For more information, see related release notes points.

Capture

Confidence Scores for Capture forms

The Forms and Classification workflow endpoints are now enriched with the Confidence Scores. Now, alongside extracted data, you will receive confidence scores of extracted fields to indicate data reliability. It helps you make accurate decisions, discard low-confidence extractions, and optimize your workflow. Some key benefits of this feature include:

• Receive confidence scores: Understand data reliability through confidence scores accompanying extracted data.
• Enhanced decision-making: Make informed choices, discard low-confidence data, and streamline workflow.
• Seamless transition: Flexibly switch between modes based on scores for optimized response evaluation.

To know more about this feature, see the Confidence score section.

Enhancements

Analyze

New tags, data sources, and metrics

The following enhancements are done in the analytics endpoints:

Enriched transactions

• New tags: Two new tags, overdraft and returned_item are added in the response for more precise data categorization. These tags replace the older tags nsf_returned_or_not_paid and nsf_paid_or_negative_balance. The nsf tag signifies transaction fees linked to Non-Sufficient Funds (NSF) incidents, the overdraft tag represents transaction fees associated with overdraft events, and the returned_item tag signifies items eligible for return due to Non-Sufficient Funds (NSF) occurrences.

  📘

  Note:

  The older tags will remain in the response temporarily to ensure a smooth transition and avoid any breaking changes.

• New data sources: Two new data sources fintech_loan_source and fintech_mca_source are incorporated to provide deeper insights into your financial information. This gives information on an applicant's loan sources and empowers underwriters to discover the lenders or MCAs the borrowers are engaged with, the amount they've received, the payments made, and the frequency of such transactions.
• Expense formula adjustment: The expense formula is revamped to exclude internal_transfer and returned_item. This adjustment ensures more accurate expense calculations.

Book summary

• New metrics: The fintech_loan_sources, fintech_mca_sources, num_fintech_loan_sources, and num_fintech_mca_sources metrics are added to provides an overview of a borrower's borrowing activity, both at the Book and Bank Account levels. It summarizes the names of sources for fintech loans and MCA, along with the counts for each source.
• Simplified metrics: To improve clarity and accuracy, we've removed all metrics related to nsf_returned_or_not_paid and nsf_paid_or_negative_balance and added new metrics that focus specifically on overdraft, returned_item, nsf_divisor, and overdraft_divisor.

Analytics (Excel)

• Added overdraft and returned_item tags to the Enriched Transactions tab in the Excel output.
• Added fintech_loan_sources, fintech_mca_sources, num_fintech_loan_sources, num_fintech_mca_sources, overdraft metrics, returned_item metrics, nsf_divisor, and overdraft_divisormetrics in the Excel output.