Planning Data vs. Filed Actuals: Why Rexfin Never Lets Them Mix

A filed annual report will tell you the audited revenue for the year. It will not tell you the monthly cadence that revenue moved through, the budget it was measured against, next year’s forecast, or how it split across business units. That detail simply does not exist in a public filing, and yet an FP&A team cannot function without it. Every planning tool eventually has to answer the same question: where does that missing detail come from, and how do you stop it from quietly becoming indistinguishable from the numbers you can actually prove?

Most tools answer badly. A budget figure gets pasted into the same grid as an audited actual, a forecast cell sits one row below a cited total, and six months later nobody remembers which was which. Rexfin answers by building the planning layer as a genuinely separate, walled track: one that can generate the FP&A breadth a CFO needs without ever being allowed to weaken the canonical atom store it sits beside.

What the planning layer actually produces

Anchored to a company’s audited annual figures, Rexfin derives the internal planning detail that filings never contain: a twelve-month cadence for the income statement and cash flow that sums exactly to the audited annual, month-end balance sheet snapshots that stay in balance and land on that same anchor, a budget, a forward forecast for the next fiscal year, segment and business-unit splits that sum back to the consolidated figure, headcount, opex detail, and the operating KPIs (margins, working-capital days, current ratio, return on equity) that FP&A work actually runs on.

None of this is invented loosely. Before anything is written out, it has to pass a reconciliation gate: months have to sum to the annual, segments have to sum to the consolidated total, and every statement has to tie. A run that fails reconciliation is not silently corrected: it fails loud instead. And it is deterministic: no model is asked to improvise a number. Given the same inputs and the same seed, the same plan comes out byte-for-byte identical every time, which means a CFO can regenerate it and get back the exact same figures they saw last week.

The wall that makes it safe

The part that matters more than the generation itself is what the planning layer is not allowed to touch. It draws only on the audited annual figures as its anchor: it cannot read from the extraction and citation path that produces Rexfin’s canonical atom store, and nothing it produces is permitted to look like that path’s output. Every planning record is explicitly tagged as unaudited and synthetic, carries no citation, and can never claim to be anchored to an actual. That tagging isn’t a note in a changelog, it is enforced, and a plan record caught smuggling a citation or a document reference fails validation rather than being quietly waved through.

This is the same logic that governs how Rexfin ranks evidence everywhere else. The trust chain grades every number a CFO sees: a cited, filed figure outranks an internal record, which outranks a plan or forecast number, and the planning layer is precisely why that bottom tier has to exist as its own category. A budget assumption is useful. It is not evidence. Rexfin’s job is to make sure it is never mistaken for evidence, no matter how many rows down the model it sits from an audited total.

That same discipline is what makes forecast baselines trustworthy: freezing a forecast and measuring drift against it only means something if everyone agrees, permanently, that the frozen version was a plan and the thing it’s compared against is an actual, never the reverse.

Why the separation is the point, not a limitation

It would be simpler to let planning figures live in the same table as cited actuals and add a label column. Rexfin doesn’t, because a label is a convention a person can ignore and a query can drop; a structural wall is not. The planning layer cannot import from the extraction path, and a plan record carrying a citation is rejected outright, not flagged for review later.

The payoff is a dataroom that can show a CFO both halves of the picture at once: cited actuals that are individually provable, sitting beside a clearly labelled plan that is openly an assumption, without either one ever bleeding into the other’s authority. That boundary is what lets Rexfin build real FP&A breadth: variance analysis, rolling forecasts, scenario work, segment and KPI detail, all the things a filing alone cannot support, without asking anyone to take a plan number on faith. For the rest of how this foundation holds together, see the inside-the-rexfin-platform hub.