LoanLogics finds persistent 11.5% error rate in US mortgage files

Audit software and automation solutions company LoanLogics announced Tuesday that 11.5% of all U.S. mortgage file content was missing or erroneous over the past 10 years, according to an analysis of internal data.

LoanLogics provides technology used to originate, manufacture, audit and service loans. Its systems process more than half of all U.S. mortgages each year. Since 2005, the company has extracted and organized nearly 16 billion data elements from unstructured sources and processed some 1.34 billion documents.

Based on its analysis of industry data and after evaluating thousands of lenders, LoanLogics estimates that inefficient systems, loan file errors and resulting delays throughout the mortgage process have translated to approximately $7.8 billion in higher costs to consumers.

LoanLogics examined data for 2014, 2019 and 2024, evaluating “doc to data” discrepancies where information claimed in a file is incorrect or missing. It also reviewed “doc to doc” discrepancies where documentation claimed to be part of a file is incorrect or missing.

“Our results show zero material improvement in loan file quality after a decade of industry investment and supposed innovation,” said Craig Riddell, executive vice president of market development at LoanLogics.

“A common challenge with recent technology investment is inappropriate application. Poor results and redundancy are the by-product of incomplete or poor training, rushed implementation, and aging integrations, leading to unexpected and costly data conflicts that necessitate manual intervention,” he added.

Notably, the combined error rate for doc-to-data and doc-to-doc transfers increased from 9.7% in 2014 to 13.3% in 2019, before declining to 11.4% in 2024.

“The spike in error rates in 2019 correlates to higher mortgage volumes across the industry. This was likely due to fluctuations in inexperienced staff brought on to deal with the increased workload,” said Roby Robertson, executive vice president of origination technology strategy at LoanLogics. “Lenders responded to the reduced volume in 2024 with reductions in workforce, leading to more experienced and knowledgeable staff, and we saw error rates come down as a result but still close to the 10-year average.

“As we continue to see new creative lending approaches emerge to serve more and more of the borrower population, it is imperative that companies work to solve their data problems with better automation and technology,” he added. “We are helping everyone from consumer-facing lenders to aftermarket loan purchasers and securitizers understand what’s broken in their files, and not merely identify issues, but also correct them.”

Why is the error rate still so high?

Despite many investments in technology and automation throughout the past decade, a high error rate signals a need for a different remedy.

“The investments that have been made into mortgage origination over the past decade have really tried to make people’s lives a little more convenient but haven’t fundamentally changed what happens,” Robertson said in an interview with HousingWire.

“The analogy I give is that a mortgage is still like a big cardboard box full of files and it’s just kind of moving down the assembly line. And that next person opens that box up and makes their assessment of what’s going on in there, and then when they say they’re done, they push that box over to the next person. That person opens the lid, and then they have to decide, do they trust what was in that box or do they not?”

At a fundamental level, Robertson said, whether a file moves down the line properly and gets assessed is up to the individual.

“That’s why the cost to manufacture a loan has skyrocketed,” he said, adding that the cost is more than $11,000, per Freddie Mac data.

Erroneous loans can also create issues when it comes to loan buybacks. Riddell said that LoanLogics can tell when a loan will contribute to the rising rate of errors.

“If a loan starts to wobble when it’s in payment issue, that’s when that file is going to get a deep dive and, coupled with performance and possibly some data flaws, that’s where some buyback activity starts to kick in,” Riddell said.

“There’s a difference between a data inconsistency and manipulation and fraud. So they have to comb through what types of errors are found.”

At the end of 2024, Robertson said, lenders began drifting outside the boundaries of traditional mortgage lending.

“At the end of 2024, they were pushing the edge of a traditional box, creating non-QM,” he said. “It was not a deliberate non-QM strategy, but deals falling outside traditional lending, going jumbo, or dealing with bad credit or high DTI to get the deal over the line. The industry was still reeling and trying to make every deal happen.”

In other words, Robertson said that lenders were simply doing what was needed to close loans. This often meant approving borrowers with higher-than-usual debt-to-income ratios, lower credit scores or loan sizes that exceeded conforming limits. Ultimately, this has resulted in more defaults in 2025.

But Robertson believes these will ease.

“Lenders have noticed [the defaults] right away. I think the data they’re getting from these portfolios is just much better than they used to. So they recognize that right away, and they’re actually shifting to doing more smart non-QM deals. They’re going after non-QM deals that make sense.”