What still needs an underwriter to decide?
Bring unresolved conflicts, policy exceptions, risk signals and reviewer notes into one place so the remaining judgment is clear instead of buried across the file.
What is left
The report is not a summary of the file. It is what the file could not settle.
Fifty findings came off this file. Forty-three are closed and stay in the file underneath. Seven are not, and those seven are the report.
- 50
- Findings off this file
- 43
- Closed behind the report
- 07
- Carried into it
The report is the remainderChoose a stage to read what it left open. Every mark is one finding; the ones lying flat are closed and stay in the file underneath, and the ones standing up are what a person still has to weigh.
- 01FinancialsA month that does not look like the others. The deposits are counted, the gap is measured, and what it means is not something a statement can answer.
- 02VerificationTwo conflicts between sources. Cevrynt shows what each record says and which record it came from, and does not pick a winner between two documents that disagree.
- 03Fraud and risk signalsA pattern worth somebody looking at the same day. Detectable in the statements; what it means depends on terms that are not in the file.
- 04PolicyThe two criteria the lender's own policy declined to settle. They arrive at the report as questions, with both lines they sit between still shown.
- 05Reviewer notesA question one reviewer left for the next. Carried forward verbatim, because a note that gets rewritten on the way to the report is not really a record of anything.
What closed means here
- A decision was takenNo
- The readings agreeYes
- It stays in the fileAlways
- Reopened by a reviewerAny time
Closed means nothing about that finding is outstanding — not that Cevrynt concluded anything. Nothing is deleted or hidden: every closed finding stays in the file underneath the report, with its source attached, and any reviewer can reopen one.
A report that restates everything makes the reader do the sorting a second time. The work is not in the forty-three — it is in putting the seven at the top, and keeping the other forty-three one click underneath them for the moment somebody wants to check.
Evidence, all the way down
Every line in the report has a floor under it.
Open a line and it descends: the finding that produced it, the reading behind the finding, and the page it was taken from. One of these four stops early, and the report says so.
- 04
- Lines opened here
- 03
- Reaching a document
- 01
- Whose floor is a person
Four lines, opened all the way downChoose a line to read what its descent is for. Each step is one layer closer to the thing the sentence rests on, and the last item stops a step early because there is no document underneath a person saying something.
Why the depth is drawn rather than asserted
Every underwriting product claims its findings are source-linked, which makes the claim worth roughly nothing on its own. What can be checked is the shape: how many layers there are, whether each one names the stage that produced it, and whether the bottom is an actual page and line or a confident-sounding sentence with nothing under it.
So the figure descends in public. Three of these four reach a document, and one does not — and the one that does not says so in its own row rather than being quietly drawn to the same depth as the others.
What a trail keeps
- The document and pageNamed
- The stage that produced itNamed
- A line with no sourceNever
- A floor that is a personSaid so
A report line with nothing under it does not get written. Where the deepest layer is a reviewer rather than a record, that is what the trail shows — attributed, dated, and in the reviewer's own words rather than paraphrased into something that sounds like a finding.
The test of an underwriting report is not whether it sounds confident. It is whether somebody who was not there can get from any sentence in it down to the page that sentence rests on — and be told plainly, in the same voice, when the floor is a person rather than a document.
What it does not contain
Six things a reader arrives expecting, and none of them are here.
A report is exactly where an underwriting platform starts emitting a score. Each row names the value you came looking for, shows the slot it would sit in, and rules that slot through.
- 06
- Things it does not issue
- 00
- Numbers of our own
- 06
- Said plainly on the page
The section a report page usually skipsChoose a row to read why it is ruled out. Each one names the value a reader arrives expecting, shows the slot it would sit in, and rules that slot through — and the middle figure above is the one that matters, because there is no number in this product that Cevrynt invented.
Why this is a section rather than a footnote
A report is where an underwriting platform starts emitting a score, because by that point in the workflow it has read everything and the temptation to summarise it as one number is enormous. The number always looks like the most valuable thing on the page and is always the least defensible, because it is the only figure nobody can trace to a document.
Ruling these out is not modesty about what the platform can do. It is the condition that makes everything else on this site checkable: if the report contained one invented number, a reader would be right to wonder which of the others were invented too.
What that leaves the report doing
- Producing a numberNever
- Carrying what is openAlways
- Naming what each is waiting onAlways
- Deciding any of itNever
None of this makes the report shorter. It makes it a different kind of document: an account of what the file showed and what remains unresolved, rather than a verdict with the working folded up inside it.
The number would have been the easiest thing on this page to build and the first thing anybody asked for. It is not here because the moment a report contains one figure nobody can trace to a document, every other figure in it becomes a matter of trust rather than a matter of record.
The decision record
The moment somebody decides, the report stops moving.
One chronology, cut once. Above the seal, everything that formed the decided report. Below it, everything that happened afterwards — kept, dated, and deliberately outside it.
- 06
- Events before the decision
- 01
- Decision, entered by a person
- 03
- After it, and outside it
D3 · 14:22 — the report is sealed
A decision is entered by a named person.
- Entered by
- Senior underwriter
- Policy in force
- v3.3
- Open findings at the time
- Seven
- Produced by Cevrynt
- None of it
Cevrynt records that a decision was entered, by whom, at what time, and against which version of the policy. What the decision was belongs to the lender — it is their call, made on their authority, and it is not ours to print on a marketing page or to generate on their behalf.
One chronology, cut onceChoose an entry to read what it did or did not change. Everything above the seal formed the decided report. Everything below it happened afterwards, sits on the file, and is deliberately not folded back into what was decided.
Why the cut is the product
A file keeps moving after it has been worked. A statement arrives late, the credit committee edits a threshold, somebody pulls the deal for a portfolio review a month on. Without a seal, the report quietly answers a different question every time it is opened — and the one question it stops being able to answer is the only one that ever gets asked in a dispute.
So the report answers what was decided, not what would be decided now. Everything after the seal is kept, dated and visible; none of it is allowed to reach back into a decision that was taken before it existed.
What the seal fixes in place
- The policy version usedFixed
- What was open at the timeFixed
- Who entered the decisionNamed
- Re-run on a later ruleNever
- Later events on the fileKept, and outside
Nothing after the seal is discarded — it is dated, visible and attached to the file. What it cannot do is change a report that was already decided, because a record that keeps agreeing with the present is not a record of anything.
Six months on, nobody asks what the rules say now. They ask what was known on day three, which version was in force, who decided, and what was still open when they did — and a report that cannot answer those four questions in that order was never really evidence of anything.
The seven
Four errands, one question, and two judgments.
Here are the seven, sorted by what unblocks them rather than by how serious they look. Severity ordering is a score with the arithmetic hidden, and it never tells anybody what to do next.
- 07
- Open findings in the report
- 04
- Waiting on paper
- 03
- Waiting on a person
Waiting on a document4 findings
Waiting on the borrower1 finding
Waiting on a reviewer's judgment2 findings
Sorted by what unblocks themChoose a finding to read what it is waiting on. Sorting these by severity would be a score wearing a different hat, so they are sorted by what actually clears them — and four of the seven turn out to be errands rather than judgments.
Why this ordering and not another
Every instinct says to rank these by how bad they are, and every ranking like that is a score with the arithmetic hidden. It would also be the least useful thing to hand somebody at nine in the morning, because severity does not tell you what to do next.
What unblocks a finding does. Four of these clear when a document turns up and need nobody's opinion at all; two need a named reviewer to weigh something; one needs a question put to the broker. That is a morning's work described accurately, rather than a queue with a number next to it.
What each finding carries
- What clears itStated
- Who it is withNamed
- The stage it came fromNamed
- A severity numberNone
No finding here is ranked against another. The lanes are uneven because the work is uneven, and the only ordering in the report is the one that tells a reader which of these they can actually do something about today.
Four errands, one question and two judgments. That is what forty-three closed findings and one sealed decision leave behind — and it is a far more useful thing to hand somebody than a number between one and a hundred that they would have had to take apart again anyway.
Reopening
Closed can be reopened. Sealed cannot be edited.
Those two promises look like they contradict each other. One request, put before and after a decision, is the clearest way to show that they do not.
- 02
- Paths a reopening can take
- 01
- Where the report is rebuilt
- 00
- Sealed reports edited
One request
A reviewer wants a closed finding reopened.
The finding was closed on day two: the deposits for three statement periods were read, structured and agreed with the statements they came from. Somebody now believes one of those periods was read against the wrong account. The request is identical in both branches below. What it does is not.
While the file is open · before day three
The finding reopens in place.
- 01The finding moves back to open, with the reviewer who reopened it and the reason recorded against it.
- 02Its earlier closed state is kept in the file's history rather than overwritten, so both readings stay visible.
- 03The report is rebuilt: forty-three closed becomes forty-two, and the open count moves from seven to eight.
- 04Nothing is contradicted, because nothing has been decided yet.
The report is rebuilt · nothing is overwritten
After the report is sealed · day three onward
The sealed report is not touched at all.
- 01The sealed report still reads seven open findings, policy v3.3, and the name against the decision. None of those numbers move.
- 02The reopening opens a new evaluation, dated, sitting below the seal with everything else that happened afterwards.
- 03Anybody reading the decision sees what was known on day three. Anybody reading the file sees that and the reopening.
- 04If the reopening changes the answer, that is a new decision with its own record — not a correction folded back into the old one.
A new evaluation below the seal · the old one stands
The two rules, and why they are not in conflict
This page makes two promises that look like they contradict each other: any reviewer can reopen a closed finding at any time, and a sealed report never changes. A reader who noticed both would be right to ask which one is actually true.
Both are, because they apply to different things. A finding is a live object on a file and stays reopenable for as long as the file exists. A report is a statement about a moment, and the moment does not come back. Reopening after a seal does not edit the sealed report — it starts a new evaluation underneath it, dated, with the old one still readable exactly as it stood.
What reopening never does
- Overwrite the earlier readingNever
- Alter a sealed reportNever
- Remove the original findingNever
- Happen without a name on itNever
Reopening is generous on purpose — the cost of making a reviewer feel they cannot revisit something is far higher than the cost of keeping two readings. What it is not allowed to do is reach backwards into a decision that was taken before the second reading existed.
A file that cannot be reopened makes people quietly wrong rather than publicly corrected. A report that can be edited after the fact is not evidence of anything. Keeping both is not a compromise between them — it is the only arrangement in which either one is worth having.
Founder-led
Send us the file that took the longest to write up.
Not the clean one. The file where the answer was obvious to whoever worked it and took two days to explain to anybody else, because the reasoning was spread across a spreadsheet, a thread and somebody's memory. That is the file worth walking through together.