1
0
The idiom, “it all comes out in the wash”, according to one of two Merriam Webster definitions, is used to say that the truth will be known in the future. The same can be said about what will come out in an effective loan quality management process. Pre close reviews can be thought of as “pretreatment” of loan file data to discover “stains” that can addressed earlier in the production cycle. Document processing is what lifts data out of documents so that loan file data and documentation can be fully compared with the result being purified data. Post close this data not only helps to improve efficiency through automation, but review accuracy as well.
Today’s audit and doc processing solutions are working in harmony to ensure “it all comes out in the wash” to deliver clean loans that will stand up to secondary market scrutiny.
Soak/Pre-wash – As discussed in a 2019 webinar, optical character recognition (OCR) is really a starting point for machine learning technologies and where the digital process for understanding your data begins. In this stage, OCR can be considered the pre-wash phase on the path to data purity. Here, OCR turns images into text or readable content. When it was the only game in town, this technology was tasked for document classification, with template-based models falling short on the document types and document variations it could identify.
(Lights, darks, denim, silk, cotton, wool, linen, viscose, rayon and so many more!)
Wash Cycle – The full context of the data is better suited to be analyzed through machine learning. Here is where the heavy lifting begins in doc processing. We’re not airing any dirty laundry by saying the mortgage industry has monstrous sized loan files and the documents that make them up are vast and highly varied. Machine learning trained systems can get their arms around these documents faster, greatly expanding the number of structured and unstructured documents that can be accurately classified through automation. This Automated Document Recognition (ADR) approach is significantly more efficient and translates into speed downstream.
Rinse – Once documents are classified, Automated Data Extraction (ADE) can be used to “rinse” the data out of them. Here, special extraction programs perform textual analysis on both structured and unstructured documents classified in the ADR phase. The result of this step is technology acceleration for identification of hundreds to thousands of data elements that can power automated loan file reviews and other business intelligence applications.
Spin Cycle – When rules-based automation takes over in this stage of data and doc processing, data errors, data inconsistencies or missing critical documents can be revealed and cleaned based on loan type or specific investor guidelines. Audit rules automation can then take over to focus auditors on exceptions so that everything comes out in the wash through the completion of post close QC reviews.
For an efficient, automated path to purified loan file data and to ensure origination blemishes are pretreated as early as possible, mortgage organizations should look to vendors with the right mix of technologies and “wash cycle” experience to train machine learning systems and automate more of the loan quality management process.
Learn more about the power of machine learning in this infographic.
