Launched in April 2020, LoanLogics IDEA™ OnDemand, was designed to improve the speed, accuracy, and affordability of in-line mortgage doc processing. At the time of launch and ever still so, this digital assistant plays an important role cutting through the immense industry volume by using automation, not people, to classify documents and extract data. All in real-time!
One of the solution’s key features that helps achieve faster throughput is its use of confidence scoring. For those not familiar, this lender configurable capability notifies users when system’s confidence in recognizing a document or data point falls out of preset range. These confidence score configurations can be set using IDEA OnDemand default ranges, by document, by data field, or globally (across all documents and/or data field).
This “manage by exception” doc processing greatly reduces processing time by instructing the automation to assign less relevance to classifying documents precisely that play a smaller role in a lender’s underwriting reviews, as an example. Lenders instead can govern the data and documents important to their business ore closely.
Because confidence scoring is not widely offered in the mortgage industry, it is important lenders understand what it is and how it differs from accuracy, a commonly used figure, when evaluating new tools.
Accuracy is how often an automated process, such as automated document recognition or automated data extraction, is correct in its resulting outcomes. Simply explained, if automated document recognition is presented with 100 documents, and it is correct in its resulting outcomes on 89 of those documents, it can be said that it is 89% accurate across that pool of 100 documents.
A result of our robust machine learning training set, built up over more than a decade of experience, IDEA OnDemand achieves 95% accuracy on many critical documents and data elements being recognized through automation alone.
Confidence, however, can be defined as the likelihood or probability that that the resulting outcomes from either automated document recognition or automated data extraction are correct. Confidence measures are more of a guide and not a guarantee that the resulting outcomes are right or wrong, but more so what is the probability that it is right or wrong.
Defining critical and non-critical documents, and data is common practice in our industry. Confidence scoring in IDEA OnDemand is a unique feature that makes it possible to granularly define the confidence levels that lenders want the system to trigger a more detailed review by staff to ensure the utmost accuracy on that mission critical information. It also helps reduce costs by eliminating the time valuable resources spend looking at documents that are insignificant to the task at hand (i.e. underwriting, due diligence review, or the like). It is also one of the top features attracting users to the product.
LoanLogics Chief Product Architecture & Innovation Officer recently spoke in more detail on the topic of confidence scoring and accuracy in the April MReport article “Removing Costs with Confidence.”
Lenders interested in learning more about confidence scoring in LoanLogics IDEA OnDemand can sign up for a try before you buy program and experience it firsthand using their own documents and staff without commitment. Request your free trial here.