The rerun loop: from a Refer to an approvable structure
How the findings are read, every lever is tried, and the next structure comes back as fields, round after round, until the loan passes.
When DU or LPA does not approve a loan, someone has to work out what to change. In many shops that is one experienced person: change something, rerun, read the new findings, try the next thing, and keep going until the loan works or the real options run out. Every difficult file waits for them.
Proxara runs that loop. This is how it works, round by round.
Read what failed, and how
The findings report is read like any other document into a fixed schema: the recommendation, the ratios, the loan amount, the LTV, the reserves, the product and the term as the AUS states them, and every message with its wording and its page. Each message is mapped to a known kind of issue. The product cannot give a reason the AUS did not give.
Two kinds of failure are treated differently, because they are different. An eligibility failure has a published threshold: a debt-to-income ratio over the limit, an LTV over the product cap, a loan amount over the county limit, reserves short of the requirement. The exact change that clears it can be computed. A risk-based Refer has no published threshold, because the AUS does not publish its risk assessment. There the product ranks the moves that reduce risk, and the AUS answers.
Every lever, tried
The levers are typed operations, not advice: lower the loan amount, pay down or pay off a liability, add or document assets, change the term, change the product, run the other AUS, add a co-borrower, switch between fixed and adjustable. Each one knows what it may change, what it needs first and which rules constrain it, and each computes its own values.
Take an example loan, Loan 2031. LPA’s first round found an LTV of 90.00% against an 85% maximum for the property. Lowering the loan amount from 360,000 to 340,000 clears it. Paying off a liability, changing the term or adding a co-borrower do not move the LTV, so those close, and the result says why.
The answer is a patch
What comes back is not a paragraph. It is the field, its current value, its proposed value, the issue it clears and the rule behind it, written into the loan and posted in Teams, where the loan officer already works. Where the AUS itself names a single change that would clear a finding, that number is shown first and attributed to the AUS.
In the shop’s own order
When several structures are valid, the order is the shop’s, not ours. The sequence its specialist already follows, this first and then that, is set once. As reruns come back, the shop’s own history reorders the choices by what actually worked there for that kind of issue. Nothing is reordered past an eligibility rule.
Round two
Proxara submits the loan again itself, through the shop’s own AUS access. The new findings are read within seconds, and the product continues from where the loan now stands, not from the first report. On Loan 2031 the second round came back Accept, and the result PDF, with every round and the page behind every figure, went into the loan file.
When no path remains, the result says which paths were tried and why each one closed, so the specialist starts where the product stopped rather than from the beginning.
The loop is where a restructure specialist’s day goes. Done in code, it runs on every file at the speed of a rerun, and the specialist is left with the files that need a person.
