When one loan can be sold twice.
Non-bank lenders raise their funding from many funders at the same time. So when one collapsed very publicly, the funders behind it discovered they'd been sold the same mortgages over and over. Nobody had spotted it, because nobody could see the whole picture.
The common response is to appoint a verification agent and have the originator hand over its funders' tapes. But the originator chooses which funders to name. Leave one off the list, and the duplicate behind it stays invisible.
None of this assumes bad faith. Most discrepancies turn out to be honest data errors, and originators and funders want the same thing. But trust is worth more when someone can actually check it. ROSA checks the data itself, every month.
The check has two parts.
Every party sends us the data it already produces, separately. We do the rest, and report the same finding back to everyone.
We find the same loan across every funder.
The originator sends us its whole book. Each funder sends its own slice. We line them up and look for any loan that's live under more than one funder with no transfer or redemption to explain it.
Match on the originator's loan ID. Those are set by the originator, and differ from one feed to the next. They can be re-issued at any time, so the same loan can wear two different numbers. It's the one identifier the data lets change freely, so we don't rely on it.
Instead, we identify a loan from a fingerprint of facts fixed when it was written. Those facts sit on every funder's tape, and they read the same on each one, because every feed describes the same original event. Same loan, same fingerprint, whatever any feed happens to call it.
And the figures on a tape have to agree with each other, so a single altered field can't hide on its own. Covering it up means changing several more in lockstep, which is its own anomaly. Tampering doesn't get easier the harder you try. It gets louder.
And a hidden funder rarely stays hidden.
Matching only works on the funders we know about, so we also keep the register of who funds whom. A funder weighing up a new line can ask us how many others already sit behind that originator.
Everyone has the same reason to ask, because nobody wants to be the funder left outside the count. So a line left off the list has a way of surfacing. The next funder to look has every incentive to bring it into view. A hand-over-your-tapes check leaves that blind spot wide open. ROSA narrows it to almost nothing.
- Funder A on file
- Funder B on file
- Funder C on file
- Funder D surfaced
A new funder asks “how many others are already here?” Anyone left off the list now has a reason to come into view.
Almost nothing
to switch on.
ROSA is a data-integrity check, not an audit. A heavyweight audit buys you complexity and a comfortable feeling. We'd rather give you the truth, and work the way your data already works. You send your standard tape; every month you get a report back.
- The full set of funders behind each originator's book.Just that one cluster, not everyone a funder funds elsewhere.
- Each party's standard data tape, sent to us separately.The file you already produce each month. Nothing reformatted.
- That's the whole list.Every month, we cross-check it all and send each party its own report.
- No access to your systems or databases.
- No access to bank accounts or money flows.
- No integration with your pipelines.
- No changes to your data tape. In fact, please don't.We want to see it as it really is. Untouched data is better evidence.
- No software to install, no onboarding project.
The lighter the touch, the truer the picture.
ROSA never needs to know who anyone is.
Our method needs no personal data. We identify a loan from non-identifying facts, never from names, addresses, or dates of birth. We never match on anything personal.
You don't have to take this on trust. It's simply how the check works.
Graded findings, plainly put.
Everyone gets the same finding and the same evidence. Findings are graded rather than a flat pass or fail, so the loud alarms stay rare and worth listening to.
Clean
Every loan accounted for across every funder. Nothing shows up twice without something to explain it.
Pre-completion overlap
The same loan shows up under two funders before completion, while there's still time to stop it.
Confirmed control failure
An overlap that's gone all the way through to completion under two funders. Raised as a confirmed failure, with the evidence attached.
A mortgage spends months in the pipeline before it completes. ROSA catches an overlap inside that window, while it can still be prevented and well before any money moves.
We're the fire alarm, not the fire brigade.
ROSA grades · you adjudicate
Every month, both sides see the same picture.
Nothing about the routine changes month to month, and that predictability is the point. The same loop, the whole book, the same report to every party.
Everyone submits
Each party sends us its standard tape, separately and securely.
We reconcile
The whole book, pipeline and completed, fingerprinted and cross-checked.
Everyone hears back
One graded report per party, in their own private folder, within the cycle.
Independent assurance, and first sight.
Confidence that the loans you're funding aren't pledged elsewhere, and the first look at any that are.
Turn your word into proof.
Make "we don't double-pledge" something you can actually show. Checked independently each month, with the ROSA mark to back it up.
ROSA answers to the relationship itself. Its independence is structural; its method is open.
- Contracted to every party on the same terms, with no stake in any loan's outcome.
- The same finding goes to everyone at once. No party can commission, suppress, or pre-view another's result.
- Every escalation carries the matched evidence, reproducible from the tapes each party submitted.