Bank data without manual re-entry: how DocuFlow shortens the path to a credit decision

At StringData we have long been developing DocuFlow, our platform for automated data extraction and document processing. Alongside continuous improvements to the module for extracting data from financial statements, our recent work has focused heavily on bank statements and transaction data.

Bank data is an important source of information when assessing the financial situation of an applicant for a loan, a lease or another financial product. Processing it by hand, however, can slow the whole process down and introduce the risk of errors. DocuFlow therefore makes it possible to process bank statements automatically and obtain the required data from them.

For clients who also want to work with transaction data taken directly from bank accounts, we have worked with CRIF to extend the solution to include data access via PSD2.

And obtaining the data is only the beginning. What matters just as much is its quality and reliability, and then the ability to turn a large volume of individual transactions into information that genuinely supports a decision.

Two routes to bank data, one goal

Bank data can enter the solution by different routes, depending on what a particular process requires.

The first route is automated processing of a bank statement using DocuFlow: the applicant supplies the statement, typically as a PDF, and the solution extracts the relevant data into a structured format.

The second option is obtaining transaction data directly from bank accounts via PSD2. We provide this route in cooperation with CRIF, whose services our solution is integrated with.

In both cases the goal is the same: high-quality structured data that can move on into downstream processes without needless manual re-entry, and help speed up the assessment of an application.

Extracting the data is not the same as trusting it

When the data comes from a bank statement, accurate extraction is only the first requirement. The figures also have to be internally consistent, and the document itself has to be credible.

DocuFlow therefore includes automated validations that help detect discrepancies in the figures or incomplete bank statements, for instance.

A further level of checking targets the authenticity of the PDF document itself. The solution can analyse its metadata and information about its origin, including the tool the file was created in. If the document carries an electronic signature, its presence and validity can be verified as well.

These mechanisms help flag potentially problematic inputs before they become the basis for further decisions. Automation is therefore not only a route to obtaining data faster, but also to better control over its quality and trustworthiness.

From transactions to information you can act on

Whether bank data comes from a statement or is obtained via PSD2, a list of transactions on its own does not yet give a clear picture of an applicant's financial situation.

The data obtained can therefore be processed further using NEOS, CRIF's analytical tool, which automatically categorises transactions — both data extracted from bank statements by DocuFlow and data obtained via PSD2.

Categorisation helps distinguish between individual types of income and expenditure, and turns a mass of transaction data into structured information suitable for further analysis. A financial institution can then assess an applicant's financial situation, and the indicators relevant to their creditworthiness, more efficiently.

Less transcription, more decision-making

The point of developing DocuFlow is to push automation further than the extraction of a document. We want the outcome to be high-quality, trustworthy data ready for use in business processes.

For bank data, that means faster processing, fewer manual interventions and better material to base decisions on. And that is the direction in which we intend to keep developing DocuFlow.