The Trust Chain: Why a Filed Number Outranks an Internal One in Rexfin

Not every number in a finance function carries the same weight. A figure from an audited, filed financial statement is a different kind of fact than a number sitting in an internal spreadsheet, which is itself a different kind of fact than a line in next year’s plan. Everyone in finance already knows this instinctively. The problem is that most software doesn’t. Once a number lands in a table or a chatbot answer, it tends to look exactly as authoritative as every other number next to it, regardless of where it actually came from.

Rexfin treats that ranking as something the product has to enforce, not something a user has to remember. Every number a CFO sees is graded: a cited, filed figure outranks an internal record, which outranks a plan or forecast number. That grading isn’t cosmetic. It’s the difference between a figure Rexfin will answer a question with directly and a figure it will only ever label as an assumption.

What actually links a number to its grade

The grading only works if it’s checkable, not asserted. Rexfin calls the underlying mechanism the trust chain, and it’s a sequence of individually verifiable links rather than one black-box confidence score: independent extraction backends have to agree on a figure, that figure has to reconcile to the total the filing itself printed, the figure has to carry a citation back to the exact document and page it came from, baseline accounting identities have to hold, and only then is the number eligible to pass through the gate that allows it into an export. Break any one link and the number doesn’t inherit the higher grade: it’s treated as unverified until it can pass the check it failed.

That last part matters more than it sounds. A confidence score can be high for a number that’s actually wrong, because confidence measures how the model feels about a value, not whether the value ties to anything real. Rexfin’s grading is deliberately not a confidence score. A figure is either backed by an unbroken chain back to a real document, or it isn’t, and the product treats those as two different categories rather than points on a continuum.

Why the ranking has to be visible, not just internal

The practical payoff is what happens when someone asks a question that spans categories, say, comparing this year’s filed revenue against next year’s planned revenue. Rexfin doesn’t quietly flatten those into two identical-looking numbers. The filed figure is presented as what it is: verified, sourced, checkable. The plan figure is presented as what it is: an assumption, however well-modeled. A reviewer, an auditor, or a board member should be able to tell at a glance which category a number belongs to without having to ask.

This is also what makes Rexfin’s answer engine willing to refuse rather than guess. If a question can only be answered by treating an internal estimate as if it were a filed fact, the honest response is to say so, or to say the question can’t be answered at the grade being asked for, not to quietly borrow the higher grade’s credibility. The same discipline shows up in the provenance drawer: clicking any number shows exactly which rung of the chain it earned and why, down to the source document.

Who this is for

This matters most to anyone who has to defend a number to someone outside the building (an auditor, a lender, a board) where “the system said so” isn’t an acceptable answer. If your current tools present a filed actual and a rough internal estimate with the same visual weight, this is the gap Rexfin is built to close. See how the full chain fits together in the Rexfin product tour.