Docs-to-digital matching allows you to automate comparing Plaid transaction data with bank statement data submitted during the loan application process. This ensures data consistency and improves efficiency for fraud detection and underwriting tasks.

🚧

This feature is in beta!

This page discusses functionality that is in beta. You may occasionally experience unannounced changes or bugs. We'd greatly appreciate your feedback on this feature and its accompanying documentation.

Step 1: Prerequisite - prepare the required data and access

Before you begin, ensure that you have the following access and data:

• Access to Analytics: Ensure that your organization has access to Ocrolus V2 Analytics. This feature requires the latest analytics endpoints, including the Book summary and Enriched transaction endpoints.

  📘

  Note

  Contact your account manager if you need assistance enabling V2 Analytics.

• Book ID: Identify the specific Book that contains previously uploaded bank statements. These statements can be in PDF, image, or scanned format.
• Plaid account and transaction data: Verify that your organization has an active Plaid account and can obtain Plaid JSON files containing the borrower's most up-to-date transaction data.

Step 2: Upload transaction data

To upload Plaid transaction data to a Book that contains bank statements, perform the following steps:

1. Use the Upload PDF to Book endpoint or the Upload Mixed Document PDF to Book endpoint to upload bank statements for underwriting the borrower.
2. Complete the underwriting and approval process as usual.

   📘

   Note

   Ocrolus will wait at least an hour for you to complete uploading documents to a book for underwriting. This waiting period ensures you have enough time to upload all necessary documents before Ocrolus begins the matching process.

   This waiting period may be adjusted over time.

3. When ready to fund, prompt the borrower to connect their bank account via Plaid.
4. Using your organization's Plaid account, obtain the bank transaction data for the same borrower. Ocrolus supports both Plaid Asset Report JSON and Plaid Transactions JSON formats.
5. Use the Upload JSON to Book endpoint to upload the retrieved Plaid JSON files to the corresponding Book. You can upload multiple JSON files to support multiple bank accounts or extended transaction timelines.

Step 3: Automatic matching

Once the JSON files are uploaded, Ocrolus automatically performs transaction matching by:

• Comparing Plaid and bank statement data: Ocrolus verifies that both Plaid transactions and bank statement data reference the same underlying bank account within the Book.
• Flagging mismatched transactions: Any discrepancies or mismatched transactions are identified and flagged for review. Notifications are sent via email or webhook for further action.

Step 4: Analyze updated insights

Once the matching process is complete, Ocrolus sends a notification summarizing the results. Notifications are delivered via one of the following methods:

• Webhooks: If no mismatches are found, the book.pacing.no_discrepancies_found event will trigger and notifications are sent confirming that the data from both sources (Bank Statements and Plaid JSON) match. If mismatches are found, the book.pacing.discrepancies_found event will trigger and notifications are sent with discrepancies for further action. To learn more about the webhook events, see Webhook events.

• Email summaries: If mismatches are found, an email notification provides an overview of the mismatched transactions for review. The email includes details on discrepancies, enabling users to take the appropriate action.

  📘

  Note

  Clients must provide the email addresses where result notifications should be sent. These email IDs should be shared with your Ocrolus Account Manager or Solutions Engineer during onboarding or integration setup. Notifications will be sent only to the specified recipients.

  

5: Review mismatch transactions

Now that Docs-to-Digital matching is complete, you can analyze the output to review mismatched transactions. This analysis can be performed using the following methods:

Review mismatched transactions via API

To analyze mismatched transactions via API, perform the following steps:

1. Access the Enriched transaction endpoint and refresh the data.

2. Review the mismatched_transactions node, which includes:

   • txn_pk: A unique identifier for the transaction.
   • txn_date: The date when the transaction occurred.
   • description: A brief detail or label associated with the transaction.
   • amount: The monetary value of the transaction.
   • uploaded_doc_pk: A unique identifier for the uploaded document associated with this transaction.
   • uploaded_doc_format: The source of the transaction data, such as BANK_STATEMENT or PLAID.
   • severity: Indicates the level of importance or risk associated with the transaction (e.g., low, medium, high).
   • reason_code: A string representing the reason or category for a flagged or mismatched transaction.
   • related_txn_pk: The unique identifier of a related transaction, used to link associated records.
   • tags: Keywords or labels assigned to the transaction for easier classification and filtering (e.g., "fintech", "fintech_loan").
   • counterparty: The name of the counterparty involved in the transaction.

   📘

   Note

   To learn more about the mismatched_transactions payload, see Enriched transaction endpoint response.

Review mismatched transactions via SMB Excel Export

If your team downloads the Excel report, you’ll find a clear, standalone view of the pacing results with severity scoring and summary insights. Here is the sample Excel report.

The Excel download includes five key sections:

• Overview: The Overview section provides the Book PK, document counts, and total transactions. It also displays the Book Severity at the top to provide you with a quick snapshot of the overall dataset and its risk level.
• Comparative Metrics: The Comparative Metrics section shows side-by-side figures from the bank statement and Plaid. It includes key financial details such as revenue, NSFs (non-sufficient funds), overdrafts, MCAs (merchant cash advances), fintech loan amounts, and other important metrics to help you compare sources easily.
• Critical Mismatches: The Critical Mismatches section highlights the total count and amount for high-risk transaction categories, including NSFs, overdrafts, crypto transactions, refunds, and gambling. This helps you quickly identify potential problem areas.
• Txn Mismatch Severity: This section summarizes the net amount in each severity category and shows how many transactions contribute to each total. It helps you understand the overall impact of mismatches by severity level.
• Mismatched Transactions: The Mismatched Transactions section provides detailed information for each mismatched transaction. It includes the transaction date and description, amount, severity, data source (Plaid or bank statement), tags, counterparty, reason code, and related transaction PK.